EP0398695A2 - Mémoire principale physiquement unique, partagée par deux ou plusieurs processeurs exécutant leurs systèmes opérationnels respectifs - Google Patents

Mémoire principale physiquement unique, partagée par deux ou plusieurs processeurs exécutant leurs systèmes opérationnels respectifs Download PDF

Info

Publication number
EP0398695A2
EP0398695A2 EP90305308A EP90305308A EP0398695A2 EP 0398695 A2 EP0398695 A2 EP 0398695A2 EP 90305308 A EP90305308 A EP 90305308A EP 90305308 A EP90305308 A EP 90305308A EP 0398695 A2 EP0398695 A2 EP 0398695A2
Authority
EP
European Patent Office
Prior art keywords
bus
data
processor
address
storage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
EP90305308A
Other languages
German (de)
English (en)
Other versions
EP0398695A3 (fr
EP0398695B1 (fr
Inventor
John Monroe Dinwiddie, Jr.
Lonnie Edward Grice
John Mario Loffredo
Kenneth Russell Sanderson
Ernest Dysart Baker
Gustavo Armando Suarez
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
International Business Machines Corp
Original Assignee
International Business Machines Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by International Business Machines Corp filed Critical International Business Machines Corp
Publication of EP0398695A2 publication Critical patent/EP0398695A2/fr
Publication of EP0398695A3 publication Critical patent/EP0398695A3/fr
Application granted granted Critical
Publication of EP0398695B1 publication Critical patent/EP0398695B1/fr
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G06F9/5016—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals the resource being the memory
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
    • G06F11/16—Error detection or correction of the data by redundancy in hardware
    • G06F11/1629—Error detection by comparing the output of redundant processing systems
    • G06F11/1641—Error detection by comparing the output of redundant processing systems where the comparison is not performed by the redundant processing components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
    • G06F11/16—Error detection or correction of the data by redundancy in hardware
    • G06F11/1658—Data re-synchronization of a redundant component, or initial sync of replacement, additional or spare unit
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
    • G06F11/16—Error detection or correction of the data by redundancy in hardware
    • G06F11/1675—Temporal synchronisation or re-synchronisation of redundant processing components
    • G06F11/1679—Temporal synchronisation or re-synchronisation of redundant processing components at clock signal level
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F15/00—Digital computers in general; Data processing equipment in general
    • G06F15/16—Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
    • G06F15/161—Computing infrastructure, e.g. computer clusters, blade chassis or hardware partitioning

Definitions

  • Fault Tolerant S/370 Module 9 Interconnected via Links, Networks General Description of Duplexed Processor Partner Units 21, 23 Coupling of S/370 and S/88 Processor Elements 85, 62
  • the present application relates to a method and means whereby a pair of central processing units (CPUs) each operating under its respective operating system share a single physical main storage unit.
  • CPUs central processing units
  • the present invention provides data processing apparatus comprising: a first data processing system having a main storage unit and operating under a first virtual operating system in conformance with a first instruction architecture; a second data processing system operating under a second operating system in conformance with a second instruction architecture and coupled to the main storage unit; means in the first data processing system for assigning a portion of the main storage for the use of the second data processing system and its operating system; and means in the first and second data processing systems, including a register defining the assigned portion of the main storage, for accessing the assigned portion of the main storage in response to instruction execution in one or both data processing systems.
  • the present invention also provides a method of operation of a data processing apparatus, the apparatus comprising a first data processing system operating under a first operating system and including a main storage unit arranged as a plurality of storage blocks, a second data processing system operating under a second operating system, and means for coupling the second data processing system to the storage unit, comprising the steps of:
  • the invention provides a method and means for capturing a section or zone of main storage from a first data processing system, including a first processing element, the main storage and I/O apparatus operated under a first operating system, for use by a second processing element having means coupling the second processing element to the main storage and operating under control of a second operating system, in a manner indiscernible to both operating systems.
  • a storage manager in the first operating system creates a list of entries, corresponding to unused blocks of storage, for allocating storage to processes.
  • An application program running in supervisor mode on the first processing element removes from the list a group of entries corresponding to a contiguous area of storage of predetermined size. Address data corresponding to said contiguous area of storage is transferred to said coupling means to permit accessing of the contiguous area by said second processing element.
  • the second processing element is given access to said contiguous area of storage, and the first processing element is given accpss to the remaining area of storage.
  • a special application program running on the first processing element (but not the first operating system) is given access to said contiguous area of storage.
  • IBM and System/370 are trademarks of the International Business Machines Corporation.
  • a preferred embodiment for implementing the present invention comprises a fault tolerant system.
  • Fault tolerant systems have typically been designed from the bottom up for fault tolerant operation.
  • the processors, storage, I/O apparatus and operating systems have been specifically tailored to provide a fault tolerant environment.
  • the breadth of their customer base, the maturity of their operating systems, the number and extent of the available user programs are not as great as those of the significantly older mainframe systems of several manufacturers such as the System/370 (S/370) system marketed by International Business Machines Corporation.
  • Today's fault tolerant data processing systems offer many advanced features that are not normally available on the older non-fault tolerant mainframe systems or that are not supported by the mainframe operating systems. Some of these features include: a single system image presented across a distributed computing network; the capability to hot plug processors and I/O controllers (remove and install cards with power on); instantaneous error detection, fault isolation and electrical removal from service of failed components without interruption to the computer user; customer replaceable units identified by remote service support; and dynamic reconfiguration resulting from component failure or adding additional devices to the system while the system is continuously operating.
  • One example of such fault tolerant systems is the System/88 (S/88) system marketed by International Business Machines Corporation.
  • Proposals for incorporating the above features into the S/370 environment and architecture might typically consist of a major rewrite of the operating system(s) and user application programs and/or new hardware developed from scratch.
  • the major rewrite of an operating system such as VM, VSE, IX370, etc. is considered by many to be a daunting task, requiring a large number of programmers and a considerable period of time. It usually takes more than five years for a complex operating system such as IBM S/370 VM or MVS to mature. Up to this time most system crashes are a result of operating system errors. Also, many years are required for users to develop proficiency in the use of an operating system. Unfortunately, once an operating system has matured and has developed a large user base, it is not a simple effort to modify the code to introduce new functions such as fault tolerance, dynamic reconfiguration, single system image, and the like.
  • the present improvement will provide a fault tolerant environment and architecture for a normally non-fault-tolerant processing system and operating system without major rewrite of the operating system.
  • a model of IBM System/88 is coupled to a model of an IBM S/370.
  • One current method of coupling distinct processors and operating systems is through some kind of communications controller added to each system, appending device drivers to the operating systems, and using some kind of communication code such as Systems Network Architecture (SNA) or OSI to transport data.
  • SNA Systems Network Architecture
  • OSI OSI
  • Layer n on one machine carries on a conversation with layer n on another machine.
  • the rules and conventions used in this conversation are collectively known as the layer n protocol.
  • the entities comprising the corresponding layers on different machines are called peer processes, and it is the peer processes that are said to communicate using the protocol.
  • the entire purpose of implementing such a structured set of protocols is to perform end-to-end transfer of data.
  • the major divisions within the OSI model can be better understood if one realizes that the user node is concerned with the delivery of data from the source application program to the recipient application program.
  • the OSI protocols art upon the data at each level to furnish frames to the network.
  • the frames are built up as the data coupled with corresponding headers applied at each OSI level.
  • These frames are then provided to the physical medium as a set of bits which are transmitted through the medium. They then undergo a reverse set of procedures to provide the data to the application program at the receiving station.
  • Fig. 1 shows a standard interconnection of two computer systems by means of a Local Area Network (LAN).
  • LAN Local Area Network
  • IBM S/370 architecture system is shown connected to an IBM System/88 architecture.
  • an application program operates through an interface with the operating system to control a processor and access an I/O channel or bus.
  • Each architecture device has a communications controller to exchange data.
  • a multi-layered protocol must be utilized to allow data to be exchanged between the corresponding application programs.
  • An alternative method to exchange data would be a coprocessor method in which the coprocessor resides on the system bus, arbitrates for the system bus, and uses the same I/O as the host processor.
  • the disadvantage of the coprocessor method is the amount of code rewrite required to support non-native (alien) host I/O.
  • Another disadvantage is that the user must be familiar with both systems architectures to switch back and forth from coprocessor to host operating systems - an unfriendly user environment.
  • a prior art fault tolerant computer system typically has a processor module containing a processing unit, a random access memory unit, peripheral control units, and a single bus structure which provides all information transfers between the several units of the module.
  • the system bus structure within each processor module includes duplicate partner buses, and each functional unit within a processor module also has a duplicate partner unit.
  • the bus structure provides operating power to units of a module and system timing signals from a main clock.
  • Fig. 2 shows in the form of a functional diagram the structure of the processor unit portion of a processor module.
  • the computer system provides fault detection at the level of each functional unit within the entire processor module. Error detectors monitor hardware operations within each unit and check information transfers between units. The detection of an error causes the processor module to isolate the unit which caused the error and to prohibit it from transferring information to other units, and the module continues operation by employing the partner of the faulty unit.
  • the memory unit is also assigned the task of checking the system bus.
  • the unit has parity checkers that test the address signals and that test the data signals on the bus structure.
  • the memory unit signals other units of the module to obey only the non-faulty bus.
  • the power supply unit for the processor module employs two power sources, each of which provides operating power to only one unit in each pair of partner units. Upon detecting a failing supply voltage, all output lines ,from the affected unit to the bus structure are clamped to ground potential to prevent a power failure from causing the transmission of faulty information to the bus structure.
  • Fig. 3 shows in the form of a functional diagram, the interconnection of paired S/370 processors with paired S/88 processors in the manner of a fault tolerant structure to enable the direct exchange of data.
  • the similarity to the prior S/88 structure (Fig. 2) is intentional but it is the unique interconnection by means of both hardware and software that establishes the operation of the preferred embodiment.
  • the S/370 processors are coupled to storage control logic and bus interface logic in addition to the S/88 type compare logic. As will be described the compare logic will function in the same manner as the care logic for the S/88 processors.
  • the S/370 processors are directly coupled and coupled through the system bus to corresponding S/88 processors.
  • the S/370 processors are coupled in pairs and the pairs are intended to be mounted on field replaceable, hot-pluggable, circuit cards. The detailed interconnections of the several drivers will be described in greater detail later.
  • the preferred embodiment interconnects plural S/370 processors for executing the same S/370 instructions concurrently under control of a S/370 operating system. These are coupled to corresponding plural S/88 processors, I/O apparatus and main storage, all executing the same S/88 instructions concurrently under control of a S/88 operating system. As will be described later means are included to asynchronously uncouple the S/88 processors from their I/O apparatus and storage, to pass S/370 I/0 commands and data from the S/370 processors to the S/88 processors while the latter are uncoupled, and to convert the commands and data to a form useable by the S/88 for later processing by the S/88 processors when they are recoupled to their I/O apparatus and main storage.
  • fault tolerant features are achieved in a preferred embodiment by coupling normally non-fault-tolerant processors such as S/370 processors in a first pair which execute the same S/370 instructions simultaneously under control of one of the S/370 operating systems. Means are provided to compare the states of various signals in one processor with those in the other processor for instantaneously detecting errors in one or both processors.
  • a second partner pair of S/370 processors with compare means are provided for executing the same S/370 instructions concurrent with the first pair and for detecting errors in the second pair.
  • Each S/370 processor is coupled to a respective S/88 processor of a fault-tolerant system such as the S/88 data processing system having first and partner second pairs of processors, S/88 I/O apparatus and S/88 main storage.
  • Each S/88 processor has associated therewith hardware coupling it to the I/O apparatus and main storage.
  • the respective S/370 and S/88 processors each have their processor buses coupled to each other by means including a bus control unit.
  • Each bus control unit includes means which interacts with an application program running on the respective S/88 processor to asynchronously uncouple the respective S;88 processor from its associated hardware and to couple it to the bus control unit (1) for the transfer of S.370 commands and data from the S/370 processor to the S/88 processor and (2) for conversion of the S,370 commands and data to commands executable by and data useable by the S/88.
  • the S/88 data processing system subsequently processes the commands and data under control of the S/88 operating system.
  • the S/88 data processing system also responds to error signals in either one of the S/370 processor pairs or in their respectively coupled S/88 processor pair to remove the coupled pairs from service and permit continued fault tolerant operation with the other coupled S/370, S/88 pairs.
  • S/370 programs are executed by the S/370 processors (with the assistance of the S 88 system for I/O operations) in a fault tolerant (FT) environment with the advantageous features of the S88 all without significant changes to the S/370 and S/88 operating systems.
  • FT fault tolerant
  • the storage management unit of the S/88 is controlled so as to assign dedicated areas in the S/88 main storage to each of the duplexed S/370 processor pairs and their operating system without knowledge by the S/88 operating system.
  • the processors of the duplexed S/370 processor pairs are coupled individually to the common bus structure of the S/88 via a storage manager apparatus and S 88 bus interface for fetching and storing S/370 instructions and data from their respective dedicated storage area.
  • the preferred embodiment provides a method and means of implementing fault tolerance in the S/370 hardware without rewriting the S/370 operating system or S/370 applications.
  • Full S/370 CPU hardware redundancy and synchronization is provided without custom designing a processor to support fault tolerance.
  • a S/370 operating system and a fault tolerant operating system, (both virtual memory systems) are run concurrently without a major rewrite of either operating system.
  • a hardware/microcode interface is provided in the preferred embodiment between peer processor pairs, each processor executing a different operating system.
  • One processor is a microcode controlled IBM S/370 engine executing an IBM Operating System (e.g., VM, VSE, IX370, etc.).
  • the second processor of the preferred embodiment is a hardware fault tolerant engine executing an operating system capable of controlling a hardware fault tolerant environment (e.g., IBM System/88), executing S/88 VOS (virtual operating system).
  • the hardware/microcode interface between the processor pairs allows the two operating systems to coexist in an environment perceived by the user as a single system environment.
  • the hardware/microcode resources (memory, system buses, disk I/O, tape, communications I/O terminals, power and enclosures) act independently of each other while each operating system handles its part of the system function.
  • the words memory, storage and store are used interchangeably herein.
  • the FT processor(s) and operating system manage error detection/isolation and recovery, dynamic reconfiguration, and I/O operations.
  • the NFT processor(s) execute native instructions without any awareness of the FT processor.
  • the FT processor appears to the NET processor as multiple I/O channels.
  • the hardware/microcode interface allows both virtual memory processors to share a common fault tolerant memory.
  • a continuous block of storage from the memory allocation table of the FT processor is assigned to each NET processor.
  • the NFT processor s dynamic address translation feature controls the block of storage that was allocated to it by the FT processor.
  • the NET processor perceives that its memory starts at address zero through the use of an offset register. Limit checking is performed to keep the NFT processor in its own storage boundaries.
  • the FT processor can access the NET storage and DMA I/O blocks of data in or out of the NET address space, whereas the NET processor is prevented from accessing storage outside its assigned address space.
  • the NET storage size can be altered by changing the configuration table.
  • Adding a new device to an existing processor and operating system generally requires hardware attachment via a bus or channel, and the writing of new device driver software for the operating system.
  • the improved "uncoupling" feature allows two distinct processors to communicate with each other without attaching one of the processors to a bus or channel and without arbitrating for bus mastership.
  • the processors communicate without significant operating system modification or the requirements of a traditional device driver. It can give to a user the image of a single system when two distinct and dissimilar processors are merged, even though each processor is executing its own native operating system.
  • This feature provides a method and means of combining the special features exhibited by a more recently developed operating system, with the users view and reliability of a mature operating system. It couples the two systems (hardware and software) together to form a new third system. It will be clear to those skilled in t[,e art that while the preferred embodiment shows a S/370 system coupled to a S/88 system any two distinct systems could be coupled. The design criteria of this concept are: little or no change to the mature operating system so that it maintains its reliability, and minimal impact to the more recently developed operating system because of the development time for code.
  • This feature involves a method combining two dissimilar systems each with its own characteristics into a third system having characteristics of both.
  • a preferred form of the method requires coupling logic between the systems that functions predominantly as a direct memory access controller (DMAC).
  • DMAC direct memory access controller
  • the main objective of this feature is to give an application program running in a fault tolerant processor (e.g., S/88 in the preferred embodiment) and layered on the fault tolerant operating system, a method of obtaining data and commands from an alien processor (e.g., S/370 in the preferred embodiment) and its operating system.
  • a fault tolerant processor e.g., S/88 in the preferred embodiment
  • an alien processor e.g., S/370 in the preferred embodiment
  • Both hardware and software defense mechanisms exist on any processor to prevent intrusion (i.e. supervisor versus user state, memory map checking, etc.).
  • Fig. 4 shows diagrammatically a S/370 processor coupled to a S/88 processor in the environment of the preferred embodiment.
  • the memory has been replaced by S/88 bus interface logic and the S/370 channel processor has been replaced by a bus adapter and bus control unit.
  • Particular attention is directed to the interconnection between the S/370 bus control unit and the S/88 processor which is shown by a double broken line.
  • This feature involves attaching the processor coupling logic to the S/88 fault tolerant processor's virtual address bus, data bus, control bus and interrupt bus structure, and not to the system bus or channel as most devices are attached.
  • the strobe line indicating that a valid address is on the fault tolerant processor' s virtual address bus is activated a few nanoseconds after the address signals are activated.
  • the coupling logic comprising the bus adapter and the bus control unit determines whether a preselected address range is presented by a S/88 application program before the strobe signal appears. If this address range is detected, the address strobe-signal is blocked from going to the S/88 fault tolerant processor hardware. This missing signal will prevent the fault tolerant hardware and operating system from knowing a machine cycle took place.
  • the fault tolerant checking logic in the hardware is isolated during this cycle and will completely miss any activity that occurs during this time. All cache, virtual address mapping logic and floating point processors on the processor bus will fail to recognize that a machine cycle has occurred. That is, all S/88 CPU functions are 'frozen,' awaiting the assertion of the Address Strobe signal by the S/88 processor.
  • the address strobe signal that was blocked from the fault tolerant processor logic is sent to the coupling logic.
  • the address strobe signal and the virtual address are used to select local storage, registers and the DMAC which are components of the coupling logic.
  • Fig. 5 shows diagrammatically the result of the detection of an interrupt from the S/370 bus control logic which is determined to be at the appropriate level and corresponding to an appropriate address.
  • the uncoupling mechanism disconnects a processor from its associated hardware and connects the processor to an alien entity for the efficient transfer of data with said entity.
  • the coupling logic has a local store which is used to queue incoming S/370 commands and store data going to and from the S/370.
  • the data and commands are moved into the local store by multiple DMA channels in the coupling logic.
  • the fault tolerant application program initializes the DMAC and services interrupts from the DMAC, which serves to notify the application program when a command has arrived or when a block of data has been received or sent.
  • the coupling logic must return data strobe acknowledge lines, prior to the clocking edge of the processor to ensure that both sides of the fault tolerant processor stay in sync.
  • the application program receives S/370 channel type commands such as Start I/O, Test I/O, etc.
  • the application program then converts each S/370 I/0 command into a fault tolerant I/O command and initiates a normal fault tolerant I/O command sequence.
  • the application program can switch the fault tolerant processor from its normal processor function to the I/0 controller function at will, and on a per cycle basis, just by the virtual address it selects.
  • two data processing systems having dissimilar instruction and memory addressing architectures are tightly coupled so as to permit one system to effectively access any part of the virtual memory space of the other system without the other system being aware of the one system's existence.
  • Special application code in the other system communicates with the one system via hardware by placing special addresses on the bus.
  • Hardware determines if the address is a special one. If it is, the strobe is blocked from being sensed by the other system's circuits, and redirected such that the other system's CPU can control special hardware, and a memory space, accessible to both systems.
  • the other system can completely control the one system when necessary, as for initialization and configuration tasks.
  • the one system cannot in any way control the other system, but may present requests for service to the other system in the following manner:
  • the latter is directed to the memory space containing the staged information and processes same to convert its format to the other system's native form. Then the application program directs the native operating system of the other system to perform native I/O operations on the converted commands and data.
  • Most current programs execute in one of two (or more) states, a supervisor state or a user state.
  • Application programs run in user state, and functions such as interrupts run in supervisor state.
  • An application attaches an I/O port then opens the port, issues an I/O request in the form of a read, write or control. At that time the processor will take a task switch. When the operating system receives an interrupt signifying an I/O completion, then the operating system will put this information into a ready queue and sort by priority for system resources.
  • the operating system reserves all interrupt vectors for its own use; none are available for new features such as an external interrupt signifying an I/O request from another machine.
  • a majority of the available interrupt vectors are actually unused, and these are set up to cause vertoring to a common error handler for 'uninitialized' or 'spurious' interrupts, as is the common practice in operating systems.
  • the preferred embodiment of this improvement replaces a subset of these otherwise unused vectors with appropriate vectors to special interrupt handlers for the S/370 coupling logic interrupts.
  • the modified S/88 Operating System is then rebound for use with the newly-integrated vectors in place.
  • the System/88 of the preferred embodiment has eight interrupt levels and uses autovectors on all levels except level 4.
  • the improvement of the present application uses one of these autovector levels, level 6, which has the next to highest priority. This level 6 is normally used by the System/88 for A/C power disturbance interrupts.
  • the logic which couples the System/370 to the System/88 presents interrupts to level 6 by ORing its interrupt requests with those of the A/C power disturbance.
  • appropriate vector numbers to the special interrupt handlers for the coupling logic interrupts are loaded into the coupling logic (some, for example, into DMAC registers) by an application program, transparent to the S/88 operating system.
  • IACK interrupt acknowledge
  • S/88 When any interrupt is received by the System/88, it initiates an interrupt acknowledge (IACK) cycle using only hardware and internal operations of the S/88 processor to process the interrupt and fetch the first interrupt handler instruction. No program instruction execution is required. However, the vector number must also be obtained and presented in a transparent fashion. This is achieved in the preferred embodiment by uncoupling the S/88 processor from its associated hardware (including the interrupt presenting mechanism for A/C power disturbances) and coupling the S/88 processor to the S/370-S/88 coupling logic when a level 6 interrupt is presented by the coupling logic.
  • the S/88 processor sets the function code and the interrupt level at its outputs and also asserts Address Strobe (AS) and Data Strobe (DS) at the beginning of the IACK cycle.
  • AS Address Strobe
  • DS Data Strobe
  • the Address Strobe is blocked from the S/88 hardware, including the A/C power disturbance interrupt mechanism, if the coupling logic interrupt presenting signal is active; and AS is sent to the coupling logic to read out the appropriate vector number, which is gated into the S/88 processor by the Data Strobe. Because the Data Strobe is blocked from the S/88 hardware, the machine cycle (IACK) is transparent to the S/88 Operating System relative to obtaining the coupling logic interrupt vector number.
  • IACK machine cycle
  • This feature couples a fault tolerant system to an alien processor and operating system that does not have code to support a fault tolerant storage, i.e. code to support removal and insertion of storage boards via hot plugging, instantaneous detection of corrupted data and its recovery if appropriate, etc.
  • This feature provides a method and means whereby two or more processors each executing different virtual operating systems can be made to share a single real storage in a manner transparent to both operating systems, and wherein one processor can access the storage space of the other processor so that data transfers between these multiple processors can occur.
  • This feature combines two user-apparent operating systems environments to give the appearance to the user of a single operating system.
  • Each operating system is a virtual operating system that normally controls its own complete real storage space.
  • This invention has only one real storage space that is shared by both processors via a common system bus. Neither operating system is substantially rewritten and neither operating system knows the other exists, or that the real storage is shared.
  • This feature uses an application program running on a first processor to search through the first operating system's storage allocation queue. When a contiguous storage space is found, large enough to satisfy the requirements of the second operating system, then this storage space is removed, by manipulating pointers, from the first operating system's storage allocation table. The first operating system no longer has use (e.g., the ability to reallocate) of this removed storage unless the application returns the storage back to the first operating system.
  • the first operating system is subservient to the second operating system from an I/O perspective and responds to the second operating system as an I/O controller.
  • the first operating system is the master of all system resources, and in the preferred embodiment is a hardware fault tolerant operating system.
  • the first operating system initially allocates and de-allocates storage (except for the storage which is "stolen" for the second operating system), and handles all associated hardware failures and recovery.
  • the objective is to combine the two operating systems without altering the operating system code to any major degree. Each operating system must believe it is controlling all of system storage, since it is a single resource being used by both processors.
  • the first operating system and its processor assume control of the system, and hardware holds the second processor in a reset condition.
  • the first operating system boots the system and determines how much real storage exists.
  • the operating system eventually organizes all storage into 4KB (4096 bytes) blocks and lists each available block in a storage allocation queue. Each 4KB block listed in the queue points to the next available 4KB block. Any storage used by the first system is either removed or added in 4KB blocks from the top of the queue; and the block pointers are appropriately adjusted.
  • the requests are satisfied by assigning from the queue a required number of 4KB blocks of real storage. When the storage is no longer needed, the blocks will be returned to the queue.
  • the first operating system executes a list of functions called module-start-up that configures the system.
  • One application that is executed by the module-start-up is a new application used to capture storage from the first operating system and allocate the storage to the second operating system.
  • This program scans the complete storage allocation list and finds a contiguous string of 4KB blocks of storage.
  • the application program then alters the pointers in the portion of the queue corresponding to the contiguous string of blocks, thereby removing a contiguous block of storage from the first operating system's memory allocation list.
  • the pointer of the 4KB block preceding the first 4KB block removed is changed to point to the 4KB block immediately following the removed contiguous string of blocks.
  • the first operating system at this point has no control or knowledge of this real memory space unless the system is rebooted or the application returns the storage pointers. It is as if the first operating system considers a segment of real storage allocated to a process running on itself and not reallocable because the blocks are removed from the table, not merely assigned to a user.
  • the removed address space is then turned over to the second operating system.
  • the second operating system then controls the storage stolen from the first operating system as if it is its own real storage, and controls the storage through its own virtual storage manager, i.e. it translates virtual addresses issued by the second system into real addresses within the assigned real storage address space.
  • An application program running on the first operating system can move I/O data into and out of the second processor s storage space, however, the second processor cannot read or write outside of its allocated space because the second operating system does not know of the additional storage. If an operating system malfunction occurs, in the second operating system, a hardware trap will prevent the second operating system from inadvertently writing in the first operating system space.
  • the amount of storage space allocated to the second operating system is defined in a table in the module-start-up program by the user. If the user wants the second processor to have 16 megabytes then he will define that in the module start up table and the application will acquire that much space from the first operating system.
  • a special SVC service call allows the application program to gain access to the supervisor region of the first operating system so that the pointers can be modified.
  • the storage is fault tolerant on the first processor; and the second processor is allowed to use fault tolerant storage and I/O from the first processor.
  • the second processor is made to be fault tolerant by replicating certain of the hardware and comparing certain of the address, data, and control lines. Using these techniques the second processor is, in fact, a fault tolerant machine even though the second operating system has no fault tolerant capabilities. More than one alien processor and operating system of the second type can be coupled to the first operating system with a separate real storage area provided for each alien processor.
  • the first operating system is that of the fault tolerant S/88 and the second operating system is one of.the S/370 operating systems and the first and second processors are S/88 and S/370 processors respectively.
  • This feature not only enables a normally non-fault- tolerant system to use a fault tolerant storage which is maintained by a fault tolerant system but also enables the non-fault-tolerant system (1) to share access to fault tolerant 1/0 apparatus maintained by the fault tolerant system and (2) to exchange data between the systems in a more efficient manner without the significant delays of a channel- to-channel coupling.
  • single system image is used to characterize computer networks in which user access to remote data and resources (e.g., printer, hard file, etc.) appears to the user to be the same as access to data and resources at the local terminal to which the user's keyboard is attached.
  • remote data and resources e.g., printer, hard file, etc.
  • the user may access a data file or resource simply by name and without having to know the object's location in the network.
  • derived single system image is introduced here as a new term, and is intended to apply to computer elements of a network which lack facilities to attach directly to a network having a single system image, but utilize hardware and software resources of that network to attach directly to same with an effective single system image.
  • loose coupling means a coupling effectuated through I/O channels of the deriving computer and the "native" computer which is part of the network.
  • Light coupling is a term presently used to describe a relationship between the deriving and native computers which is established through special hardware allowing each to communicate with the other on a direct basis (i. e., without using existing 1/0 channels of either).
  • Transparent tight coupling involves the adaptation of the coupling hardware to enable each computer (the deriving and native computers) to utilize resources of the other computer in a manner such that the operating system of each computer is unaware of such utilization.
  • Transparent tight coupling forms a basis for achieving cost and performance advantages in the coupled network.
  • the cost of the coupling hardware notwithstanding complexity of design, should be more than offset by the savings realized by avoiding the extensive modifications of operating system software which otherwise would be needed. Performance advantages flow from faster connections due to the direct coupling and reduced bandwidth interference at the coupling interface.
  • network as used in this section is more restricted than the currently prevalent concept of a network which is a larger international teleprocessing/satellite connection scheme to which many dissimilar machine types may connect if in conformance to some specific protocol. Rather “network” is used in this section to apply to a connected complex of System/88 processors or alternatively to a connected complex of other processors having the characteristics of a single system image.
  • the term 'single system image refers to the application program's view of the syytem and environment.
  • 'System in this context, means the combined hardware (CPU complex) and software (Operating System and its utilities) to which the application programmer directs his instructions.
  • 'Environment' means all I/O devices and other connected facilities which are addressable by the Operating System and thus accessible indirectly by the programmer, through service requests to the Operating System.
  • a truly single, free-standing computer with its Operating System then, must provide a SINGLE-SYSTEM IMAGE to the programmer. It is only when we want to connect multiple systems together in order to share I/O devices and distribute processing that this 'image' seen by the programmer begins to change; the ordinary interconnection of two machines via teleprocessing lines (or even cables) forces the programmer to understand - and learn to handle - the dual environment, in order to take advantage of the expanded facilities.
  • the System/88 original design included the means to simplify this situation and provide the SINGLE-SYSTEM IMAGE to the programmer, i.e., the HSDI connection between MODULES, and HSDI drive software within the Operating System in each MODULE.
  • the HSDI connection between MODULES, and HSDI drive software within the Operating System in each MODULE.
  • each of the two Operating Systems 'know about' the entire environment, and can access facilities across the HSDI without the active intervention of the 'other' Operating System.
  • the reduction in communications overhead is considerable.
  • a large number of MODULEs of various sizes and model types can be interconnected via HSDI to create a system complex that appears to the programmer as one (expandable) environment.
  • This product in application program can be stored on one disk in this system complex, executed in any of the CPUs in the complex, controlled or monitored from essentially any of the terminals of the complex, and can transfer data to and from any of the I/O devices of the complex, all without any special programming considerations and with improved execution efficiency over the older methods.
  • the operating system and its various features and facilities are written in such a way as to natively assume the distributed environment and operate within that environment with the user having no need to be concerned with or have control over where the various entities (utilities, applications, data, language processors, etc.) reside.
  • the key to making all of this possible is the enforced rule that each OBJECT must have a unique name; and this rule easily extends to the entire system complex since the most basic name- qualifier is the MODULE name, which itself must be unique within the complex. Therefore, locating any OBJECT in the entire complex is as simple as correctly naming it. Naming an OBJECT is in turn simplified for the programmer by the provision of LINKs, which allow the use of very short alias pointers to (substitute names for) OBJECTS with very long and complicated names.
  • a plurality of S/370 processors are coupled to S/88 processors in such a manner as to provide for the S/370 processor users at least some aspects of the S/88 single system image features. This, even though the S/370 processors and operating systems do not provide these features.
  • One or more S/370 processors are provided within the S/88 MODULE.
  • a S/88 processor is uniquely coupled to each S/370 processor.
  • each S/370 processor is replicated and controlled by S/88 software for fault- tolerant operation.
  • Each S/370 processor uses the fault-tolerant S/88 system complex to completely provide the S/370 main storage, and emulated S/370 I/O Channel(s) and I/O device(s).
  • the S/370s have no main memory, channels, or I/O devices which are not part of the S/88, and all of these facilities are fault-tolerant by design.
  • each S/370 processor is assigned a dedicated contiguous block of 1 to 16 megabytes of main storage from the S/88 pool; this block is removed from the configuration tables of the S/88 so that the S/88 Operating System cannot access it, even inadvertently.
  • Fault-tolerant hardware registers hold the storage block pointer for each S/370, so that the S/370 has no means to access any main storage other than that assigned to it.
  • the result is an entirely conventional, single-system view of its main memory by the S/370; the fault-tolerant aspect of the memory is completely transparent.
  • An application program (EXEC370) in the S/A8 emulates S/370 Channel(s) and I/O device(s) using actual S/88 devices and S/88 Operating System calls. It has the SINGLE-SYSTEM IMAGE view of the S/88 complex, since it is an application program; thus this view is extended to the entire S/370 'pseudo-channel.'
  • connection technique is relatively simple and quick dynamic reconfigurability of each S/370.
  • the channel 'window' is two-way, and the S/88 control program EXEC370 is on the other side of it; EXEC370 has full capability to stop, reset, reinitialize, reconfigure, and restart the S/370 CPU.
  • S/88 1/0 and Operating System transparent emulation of S/370 1/0 facilities using other facilities which possess the SINGLE-SYSTEM IMAGE attribute (S/88 1/0 and Operating System)
  • the S/370 therefore has been provided with object location independence. Its users may access a data file or other resource by name, a name assigned to it in the S/88 operating system directory. The user need not know the location of thB data file in the complex of S/370-S/88 modules.
  • S/370 I/O commands issued by one S/370 processing unit in one module 9 are processed by an associated S/88 processing unit tightly coupled to the S/370 processing unit in the same module (or by other S/88 processing units interconnected in the module 9 and controlled by the same copy of the S/88 virtual operating system which supports multiprocessing) to access data files and the like resident in the same or other connected modules. It may return the accessed files to the requesting S/370 processing unit or send them to other modules, for example, to merge with other files.
  • the functions of two virtual operating systems are merged into one physical system.
  • the S/88 processor runs the S/88 OS and handles the fault tolerant aspects of the system.
  • one or more S/370 processors are plugged into the S/88 rack and are allocated by the S/88 OS anywhere from 1 to 16 megabytes of contiguous memory per S/370 processor.
  • Each S/370 virtual operating system thinks its memory allocation starts at address 0 and it manages its memory through normal S/370 dynamic memory allocation and paging techniques.
  • the S/370 is limit checked to prevent the S/370 from accessing S/88 memory space.
  • the S/88 must access the S/370 address space since the S/88 must move I/O data into the S/370 I/O buffers.
  • the S/88 Operating System is the master over all system hardware and I/O devices.
  • the peer processor pairs execute their respective Operating Systems in a single system environment without significant rewriting of either operating system.
  • the IBM System/88 marketed by International Business Machines Corp. is described generally in the IBM System/88 Digest, Second Edition, published in 1986 and other available S/88 customer publications.
  • the System/88 computer system including module 10, Fig. 6A is a high availability system designed to meet the needs of customers who require highly reliable online processing.
  • System/88 combines a duplexed hardware architecture with sophisticated operating system software to provide a fault tolerant system.
  • the System/88 also provides horizontal growth through the attachment of multiple System/88 modules 10a, 10b, 10c, through the System/88 high speed data interconnections (HSDls), Fig. 6B, and modules 10d-g through the System/88 Network, Fig. 6C.
  • HSDls System/88 high speed data interconnections
  • the System/88 is designed to detect a component failure when and where it occurs, and to prevent errors and interruptions caused by such failures from being introduced into the system. Since fault tolerance is a part of the System/88 hardware design, it does not require programming by the application developer. Fault tolerance is accomplished with no software overhead or performance degradation.
  • the System/88 achieves fault tolerance through the duplication of major components, including processors, direct access storage devices (DASDs) or disks, memory, and controllers. If a duplexed component fails, its duplexed partner automatically continues processing and the system remains available to the end users. Duplicate power supplies with battery backup for memory retention during a short-term power failure are also provided. System/88 and its software products offer ease of expansion, the sharing of resources among users, and solutions to complex requirements while maintaining a single system image to the end user.
  • DASDs direct access storage devices
  • a single system image is a distributed processing environment consisting of many processors, each with its own files and 1/0, interconnected via a network or LAN, that presents to the user the impression he is logged on to a single machine.
  • the operating system allows the user to converse from one machine to another just by changing a directory.
  • the System/88 processing capacity can be expanded while the System/88 is running and while maintaining a single-system image to the end user.
  • Horizontal growth is accomplished by combining multiple processing modules into systems using the System/88 HSDI, and combining multiple systems into a network using the System/88 Network.
  • a System/88 processing module is a complete, stand-alone computer as seen in Fig. 6A of the drawings.
  • a System/88 system is either a single module or a group of modules connected in a local network with the IBM HSDI as seen in Fig. 68.
  • the System/88 Network using remote transmission facilities, is the facility used to interconnect multiple systems to form a single-system image to the end user. Two or more systems can be interconnected by communications lines to form a long haul network. This connection may be through a direct cable, a leased telephone line, or an X.25 network.
  • the System/88 Network detects references to remote resources and routes messages between modules and systems completely transparent to the user.
  • Hot pluggability allows many hardware replacements to be done without interrupting system operation.
  • the System/88 takes a failing component out of service, continuing service with its duplexed partner, and lights an indicator on the failing component - all without operator intervention.
  • the customer or service personnel can remove and replace a failed duplexed board while processing continues.
  • the benefits to a customer include timely repair and reduced maintenance costs.
  • System/88 is a fault-tolerant, continuous operation machine, there are times when machine operation will need to be stopped. Some examples of this are to upgrade the System/88 Operating System, to change the hardware configuration (add main storage), or to perform certain service procedures.
  • the duplexed System/88 components and the System/88 software help maintain data integrity.
  • the System/88 detects a failure or transient error at the point of failure and does not propagate it throughout the application or data. Data is protected from corruption and system integrity is maintained.
  • Each component contains its own error-detection logic and diagnostics. The error-detection logic compares the results of parallel operations at every machine cycle.
  • the system detects a component malfunction, that component is automatically removed from service. Processing continues on the duplexed partner while the failed component is checked by internal diagnostics. The error-detection functions will automatically run diagnostics on a failing component removed from service while processing continues on its duplexed partner. If the diagnostics determine that certain components need to be replaced, the System/88 can automatically call a support center to report the problem. The customer benefits from quick repairs and low maintenance costs.
  • the System/88 is based generally upon processor systems of the type described in detail in U.S. Patent 4,453,215, entitled “Central Processing Apparatus for Fault Tolerant Computing", issued June 5, 1984 to Robert Reid and related U.S. patents no. 4,486,826, 4,597,084, 4,654,857, 4,750,177 and 4,816,990. Portions of patent 4,453,215 are shown diagrammatically in Figs. 7 and 8 of the present application.
  • This computer system of Figs. 7 and 8 of the present application has a processor module 10 with a processing unit 12, a random access storage unit 16, peripheral control units 20, 24, 32, and a single bus structure 30 which provides all information transfers between the several units of the module.
  • the bus structure within each processor module includes duplicate partner buses A, B, and each functional unit 12, 16, 20, 24, 32 has an identical partner unit.
  • Each unit, other than control units which operate with asynchronous peripheral devices, normally operates in lock-step synchronism with its partner unit.
  • the two partner memory units 16, 18 of a processor module normally both drive the two partner buses A, B, and are both driven by the bus structure 30, in full synchronism.
  • the computer system provides fault detection at the level of each functional unit within a processor module.
  • error detectors monitor hardware operations with in each unit and check information transfers between the units.
  • the detection of an error causes the processor module to isolate the bus or unit which caused the error from transferring information to other units, and the module continues operation.
  • the continued operation employs the partner of the faulty bus or unit. Where the error detection precedes an information transfer, the continued operation can execute the transfer at the same time it would have occurred in the absence of the fault. Where the error detection coincides with an information transfer, the continued operation can repeat the transfer.
  • the computer system can effect the foregoing fault detection and remedial action rapidly, i.e. within a fraction of an operating cycle.
  • the computer system has at most only a single information transfer that is of questionable validity and which requires repeating to ensure total data validity.
  • processor module has significant hardware redundancy to provide fault-tolerant operation, a module that has no duplicate units is nevertheless fully operational.
  • the functional unit redundancy enables the module to continue operating in the event of a fault in any unit.
  • all units of a processor module operate continuously, and with selected synchronism, in the absence of any detected fault.
  • that unit Upon detection of an error-manifesting fault in any unit, that unit is isolated and placed off-line so that it cannot transfer information to other units of the module.
  • the partner of the off-line unit continues operating, normally with essentially no interruption.
  • each unit within a processor module generally has a duplicate of hardware which is involved in a data transfer.
  • the purpose of this duplication, within a functional unit, is to test, independently of the other units, for faults within each unit.
  • Other structure within each unit of a module, including the error detection structure, is in general not duplicated.
  • the common bus structure which serves all units of a processor module preferably employs a combination of the foregoing two levels of duplication and has three sets of conductors that form an A bus, a B bus that duplicates the A bus, and an X bus.
  • the A and B buses each carry an identical set of cycle- definition, address, data, parity and other signals that can be compared to warn of erroneous information transfer between units.
  • the conductors of the X bus which are not duplicated, in general carry module-wide and other operating signals such as timing, error conditions, and electrical power.
  • An additional C bus is provided for local communication between partnered units.
  • a processor module detects and locates a fault by a combination of techniques within each functional unit including comparing the operation of duplicated sections of the unit, the use of parity and further error checking and correcting codes, and by monitoring operating parameters such as supply voltages.
  • Each central processing unit has two redundant processing sections and, if the comparison is invalid, isolates the processing unit from transferring information to the bus structure.
  • Each processing unit also has a stage for providing virtual memory operation which is not duplicated. Rather, the processing unit employs parity techniques to detect a fault in this stage.
  • the random access memory unit 16 is arranged with two non-redundant memory sections, each of which is arranged for the storage of different bytes of a memory word.
  • the unit detects a fault both in each memory section and in the composite of the two sections, with an error-correcting code. Again, the error detector disables the memory unit from transferring potentially erroneous information onto the bus structure and hence to other units.
  • the memory unit 16 is also assigned the task of checking the duplicated bus conductors, i.e. the A bus and the B bus.
  • the unit has parity checkers that test the address signals and that test the data signals on the bus structure.
  • a comparator compares all signals on the A bus with all signals on the B bus. Upon determining in this manner that either bus is faulty, the memory unit signals other units of the module, by way of the X bus, to obey only the non-faulty bus.
  • Peripheral control units for a processor module employ a bus interface section for connection with the common bus structure, duplicate control sections termed “drive” and “check”, and a peripheral interface section that communicates between the control sections and the peripheral input/output devices which the unit serves.
  • the bus interface section feeds input signals to the drive and check control sections from the A bus and/or the B bus, tests for logical errors in certain input signals from the bus structure, and tests the identity of signals output from the drive and check channels.
  • the drive control section in each peripheral control unit provides control, address, status, and data manipulating function6 appropriate for the 1/0 device which the unit serves.
  • the check control section of the unit is essentially identical for the purpose of checking the drive control section.
  • the peripheral interface section of each control unit includes a combination of parity and comparator devices for testing signals which pass between the control unit and the peripheral devices for errors.
  • a peripheral control unit which operates with a synchronous I/O device, such as a communication control unit 24, operates in lock-step synchronism with its partner unit.
  • the partnered disk control units 20,22 operate with different non-synchronized disk memories and accordingly operate with limited synchronism.
  • the partner disk control units 20, 22 perform write operations concurrently but not in precise synchronism inasmuch as the disk memories operate asynchronously of one another.
  • the control unit 32 and its partner also typically operate with this limited degree of synchronism.
  • Tile power supply unit for a module employs two bulk power supplies, each of which provides operating power to only one unit in each pair of partner units.
  • one bulk supply feeds one duplicated portion of the bus structure, one of two partner central processing units, one of two partner memory units, and one unit in each pair of peripheral control units.
  • the bulk supplies also provide electrical power for non- duplicated units of the processor module.
  • Each unit of the module has a power supply stage which receives operating power from one bulk supply and in turn develops the operating voltages which that unit requires. This power stage in addition monitors the supply voltages. Upon detecting a failing supply voltage, the power stage produces a signal that clamps to ground potential all output lines from that unit to the bus structure. This action precludes a power failure at any unit from causing the transmission of faulty information to the bus structure.
  • Some units of the processor module execute each information transfer with an operating cycle that includes an error-detecting timing phase prior to the actual information transfer.
  • a unit which provides this operation e.g. a control unit fo a peripheral device, thus tests for a fault condition prior to effecting an information transfer.
  • the unit inhibits the information transfer in the event a fault is detected.
  • the module can continue operation - without interruption or delay - and effect the information transfer from the non-inhibited partner unit.
  • Other units of the processor module execute each information transfer concurrently with the error detection pertinent to that transfer.
  • the unit In the event a fault is detected, the unit immediately produces a signal which alerts other processing units to disregard the immediately preceding information transfer.
  • the processor module can repeat the information transfer from the partner of the unit which reported a fault condition. This manner of operation produces optimum operating speed in that each information transfer is executed without delay for the purpose of error detection. A delay only arises in the relatively few instances where a fault is detected.
  • a bus arbitration means is provided to determine which unit gains access to the system bus when multiple units are requesting access.
  • the Fault Tolerant S/370 Module 9 Interconnected via HSOls, Networks
  • Fig. 7 illustrates in the portion above prior art module 10, the interconnection of S/370 and S 88 duplexed processor pairs (partner units) 21, 23 which, when substituted for duplexed S/88 units 12, 14 m module 10, creates a new and unique S/370 module 9.
  • unique modules 9 are interconnected by S/88 HSOls and networks in a manner similar to that shown in Figs. 68, 6C for modules 10, they create a S/370 complex (rather than a S/88 complex) with the S/88 features of fault tolerance, single system image. hot pluggability, 1/0 load sharing among multiple S/88 processing units within the same module, etc.
  • S/370 processors in partner units 21, 23 of the unique modules 9 execute S,370 instructions under control of their respective S/370 operating system; the interconnected S/88 processors perform all of the S/370 I/O operations in conjunction with their respective S/88 storage and S/88 peripheral units under control of the S/88 operating system in conjunction with a S/88 application program.
  • S/370 - S/88 processor partner units 25, 27 and 29, 31 can be incorporated within the new module 9 to permit a S/370 plural processor environment within the unique module 9.
  • the S/370 processors within the partner units 21, 23 and 25, 27 and 29, 31 may each operate under a different S/370 operating system per partner-pair.
  • Fig. 8 illustrates a preferred form of interconnecting S/370 and S/88 processors within the unit 21.
  • the lower portion of unit 21 comprises a central processor 12 essentially identical to processor 12 of the US patent number 4,453,215 except for the use of a single processor element in each of the pair of processor elements 60, 62.
  • dual processors were provided at 60 and at 62 to execute respectively user code and operating system code.
  • both functions are performed by a single microprocessor, preferably a Motorola MC68020 Microprocessor described in the MC68020 Users Manual, Third Edition (ISBN-0-13-567017-9) published by Motorola, copyright 1989, 1988.
  • each processor element (PE)60 and 62 preferably comprises a Motorola 68020 microprocessor.
  • Multiplexors 61, 63 connect processor elements 60, 62 to the bus structure 30 by way of address/data control A and B buses and transceivers 12e in a manner described in detail in the above-mentioned patent.
  • Local control 64, 66 and a virtual storage map 12c are provided for elements 60, 62.
  • a comparator 12f checks for error-producing faults by comparing signals on control, data and address lines to and from the bus 30 and the processor elements 60, 62.
  • FIG. 8 illustrates a preferred form of connecting a pair of S/370 processing elements 85, 87 to the S/88 bus structure 30 and to the S/88 processing elements 60, 62.
  • the processing elements 85, 87 are connected to the bus structure 30 via multiplexors 71, 73 and transceivers 13 in a manner logically similar to that in which elements 60, 62 are coupled to the bus structure 30.
  • a compare circuit 15 (described more fully in Figs. 32A, B), clamp circuits 77 and 79 and common controls 75 are provided and operate in a manner similar to corresponding components in unit 12.
  • the control circuit 86 is coupled to the S/88 interrupt mechanism of processing elements 60, 62.
  • the S/370 processors 85, 87 and their related hardware use the S/88 to process error handling and recovery.
  • the common control circuit 75 is coupled to the common control circuit 86 via line 95 to permit the latter to handle errors detected by compare circuit 15.
  • This coupling line 95 also permits common controls 75 and 86 to take both of their respective processor pairs 85, 87 and 60, 62 off line in the event of an error in either processor pair.
  • a preferred form of the S/370 processing units in unit 21 includes the central processing elements 85, 87 storage management units 81, 83 and processor-to-processor (e.g. S/370 to S/88) interfaces 89, 91.
  • the storage management units 81, 83 couple processing elements 85, 87 to S/88 main storage 16 via multiplexors 71, 73 transceivers 13 and bus structure 30.
  • Interfaces 89, 91 couple the processor buses of the S/370 processing elements 85, 87 respectively to the processor buses of the S/88 processing elements 62, 60.
  • the partner processor unit 23 is identical to processor unit 21. It will be remembered relative to the above description that the two processing elements 60, 62 in unit 21 and the corresponding two elements (not shown) in unit 23 all normally operate in lock-step with each other to simultaneously execute identical instructions under control of the same S/88 operating system.
  • processing elements 85, 87 in unit 21 and their corresponding elements (not shown) in unit 23 operate in lock-step with each other to simultaneously execute identical instructions under control of the same S/370 operating system.
  • Figs. 9A and 9B show one form of physical packaging for the S/370 and S/88 components for the processor unit 21 of Fig. 8.
  • the S/370 components including the paired processing elements 85, 87 are mounted on one board 101 and the S/88 components including the paired processing elements 60, 62 are mounted on another board 102.
  • the two boards 101 and 102 are rigidly affixed to each other to form a sandwich pair 103 and are adapted for insertion into two slots of the back panel (not shown) of the module 9, conventional back panel wiring couples the components on the boards 101 and 102 to each other and to the bus structure 30 as illustrated in Fig. 8 and as described in US patent number 4,453,215.
  • Fig. 10 is used to illustrate a preferred form of the mapping of the S/88 virtual storage to real storage 16 by a storage management unit 105 for one module 9.
  • the virtual address space 106 is divided into S/88 operating system space 107 and user application space 108.
  • Within the space 107 is an area 109 (addresses 007EOOOO to 007EFFFF) reserved for hardware and code used to couple each S/370 processor element to a respective S/88 processor element in a processor unit such as 21.
  • the address space 109 is made transparent to the S/88 operating system during normal system processing. The use of this space 109 will be described in detail below.
  • the storage management unit 105 assigns within the S/88 main storage unit 16 a S/370 main storage area for each set of four S/370 processor elements in partnered units such as 21 and 23.
  • partnered units such as 21 and 23.
  • three S/370 main storage areas 162, 163 and 164 are provided for partner units 21, 23 and 25, 27 and 29, 31 respectively.
  • the S/88 processor elements within the partner units access the remaining parts of the storage unit 16 in the manner described in the US patent number 4,453,215.
  • the S/370 storage areas 162-164 are assigned, as will be described later, in a manner such that the S/88 operating system does not know that these areas have been "stolen” and are not reassignable to S/88 users by the storage management unit unless returned to the S/88 space. Since the S/370 systems are virtual systems, they access their respective main storage area via address translation. The partner S/88 main storage unit 18 requires identical S/370 main storage areas (not shown). Each S/370 processor element can access only its respective S/370 main storage area and produces an error signal if it attempts to access the S/88 main storage spice. Each S/88 processor element, however, can access (or direct the access to) the S/370 main storage area of its respective S/370 processor element during S/370 I/O operations when the S/88 processor element acts as an 1/0 controller for its S/370 processor element.
  • Fig. 8 illustrates diagrammatically the provision of four S/370 processor elements such as 85, two in each of the partner units 21, 23 and four S/88 processor elements such as 62, two in each unit 21, 23 coupled such that all S/370 processor elements concurrently execute identical S/370 instructions and all S/88 processor elements concurrently execute identical S/88 instructions.
  • all four S/370 processor elements act as one S/370 processing unit insofar as program execution is concerned.
  • all four S/88 processor elements act as one S/88 processing unit.
  • FIG. 11 shows the processor element 85 coupled to the system bus 30 and S/88 storage 16 by way of a first path including its processor bus 170, and a S/370 storage management unit 81.
  • PE85 is shown coupled to the processor bus 161 of PE62 by way of a second path including processor element to processor element interface 89.
  • PE85 uses the first path during S/370 program execution to fetch (and store) data and instructions from its assigned S/370 main storage area 162 in store 16.
  • PE62 performs S/370 I/O operations for PE85 over the second path including interface 89.
  • a S/370 chip set 150 (Fig. 11) includes individual functional chips for the processor element 85, a clock 152, a cache controller 153 with a directory look aside table (DLAT) 341, a bus adapter 154, an optional floating point coprocessor element 151 and a control store 171 for storing a set of microcode which supports the S/370 architecture.
  • This S/370 chip set may be adapted to be operated by any of the existing S/370 operating systems (such as VSE/SP, VM/SP, IX/370 etc.) marketed by International Business Machines Corporation.
  • the cache controller 153 together with a storage control interface (STCI) 155 form the S/370 storage management unit 81.
  • the bus adapter 154 and a bus control unit (BCU) 156 comprise the PE to FE interface 89.
  • each of the S/370 CPU's such as PE85 is a 32 bit microprocessor having 32 bit data flow, a 32 bit arithmetic/logic unit (ALU), 32 bit registers in a three port data local store, and an 8 byte S/370 instruction buffer.
  • S/370 instructions are executed either in hardware or are interpreted by micro instructions.
  • the chip 153 provides cache storage for S/370 program instructions and data together with associated storage control functions. The chip 153 handles all storage requests that are issued from the PE85 as it executes its program instructions. The chip 153 also handles requests from the bus adapter 154 when transferring I/O data.
  • the bus adapter 154 and BCU 156 provide logic end control to directly (or tightly) interconnect the internal S/370 processor bus 170 to the S/88 processor bus 161 during input/output operations.
  • the BCU 156 is the primary mechanism for directly coupling the processor buses of PF85 and PE62 to each other. It is the hardware mechanism which interacts with the S/88 processor element 62 when PE62 is "uncoupled" from its associated system hardware for the transfer of data and commands between PE62 and PE85 as will be described later.
  • the clock chip 152 (Fig. 12) uses centralized logic for clock signal generation and applies appropriate clock signals individually to each of the other chips 85, 151, 153 and 154.
  • the clock 152 is in turn controlled by clock signals from the System/88 bus 30 to synchronize both the S/370 PE85 and the S/88 PE62.
  • this interface is handled by the STCI logic 155 which must communicate between the S/370 cache controller 153 and the S/88 system bus 30.
  • the non-fault- tolerant hardware must be replicated on the board as shown in Fig. 8 to produce a 'check' and 'drive' logic which are capable of running in lock-step with each other and with a partner unit.
  • the 'single' CPU consisting of system components on boards 101 and 102, must run in lock-step with its respective duplexed partner unit.
  • the task of implementing the above requirements while maintaining optimal performance and functionality involves the synchronization of separate clock sources.
  • the S/88 system clock 38 (Fig. 7) is received by all devices attached to the common bus structure 30, and two S/88 clock cycles are defined per bus 30 cycle.
  • This system clock 38 ensures synchronous communication on the bus and may be used by individual processors/controllers to develop internal clock frequency sources based on the system clock.
  • the S/370 hardware utilizes an oscillator input into the S/370 clock chip 152, which then generates a set of unique clocks to each of the other S/370 chips 85, 151, 153, 154, 155.
  • This clock chip 152 has inherent delay which can vary based on various parameters such as operating temperature, manufacturing variations, etc. This delay variation may be unacceptable in both maintaining lock-step synchronization between redundant check and drive logic, as well as in maintaining full pipelining capability between the STCI 155 and the bus structure 30.
  • the preferred embodiment utilizes redundant clock synchronization (sync) logic 158 (and 158a not shown, for the paired S/370 processor unit) to allow both processor check and drive sides of a board 101 to run in lock-step after a reset (i.e., power-on-reset or other), while synchronizing the S/370 processor cycle with the S/88 bus 30 cycle.
  • Clock signals from the S/88 clock 38 are applied via bus structure 30 to the sync logic 158 and to the STCI 155, for S/88-S/370 synchronization and for accessing the main storage via system bus 30.
  • This synchronization is accomplished in the clock sync logic 158 by first multiplying the S/88 clock to achieve the desired S/370 oscillator input frequency into the S/370 clock chip 152. In this case it is twice the frequency of the S/88 and S/370 clock cycles. Secondly, a feedback pulse on line 159 representing the beginning of the S/370 cycle, is sampled with S/88 clocks representing the leading and trailing edges of a period one register latch delay greater than the S/370 oscillator input clock period, which itself is equal to a S/88 half-cycle period.
  • the S/370 processor cycle is assured to start within a S/88 half-cycle period of the start of the S/88 clock period. All transfer timings between the bus structure 30 and S/370 cache controller 153 thus assume the worst case delay for this half-cycle.
  • the comparator logic 15 is only fed by lines sampled with S/88 clocks, ensuring synchronization of "broken" logic 403 (Fig. 32) with the accompanying S/88 processor board 102. Therefore, although the check and drive S/370 hardware may actually be slightly out of sync due to delay variations in their respective clock generation logic, both sides will run in lock-step relative to the current S/88 clock 38 common to bus structure 30, and never more than a half-cycle after the start of the S/88 clock cycle.
  • the sync logic 158 continually monitors the S/370 clock feedback on line 159 to ensure no drifting beyond the half-cycle period.
  • a maximum of one bus 30 cycle is required in the preferred embodiment to bring both sides into sync during any system reset; however, any drift in total delay outside of reset, which causes one side to 'extend' its S/370 clocks, will result in a board "broken" condition, i.e., a fault.
  • Fig. 12 shows the arrangement of Fig. 11 in greater detail.
  • the S/370 control store 171 is shown connected to PE85.
  • the control store 171 in the preferred embodiment consists of 16KB of random access storage for storing micro instructions which control the execution of program instructions and 1/0 operations within PE85.
  • the control store 171 also includes therein a 64B block 186 (Fig. 29) which is used as a buffer to hold transient micro code loaded on a demand basis from an internal object area (IOA) 187 (Fig. 28) which is part of the S/370 dedicated storage 162 within the main storage unit 16.
  • IOA internal object area
  • PE62 has associated therewith hardware including a floating point processor 172, a cache 173, a microcode storage unit 174 which is used to store coupling microcode referred to as ETIO herein. Both the microcode and an application program stored in cache 173, as will be seen below, are used for controlling PE62 and the BCU logic 156 to perform I/O operations for PE85.
  • the PE62 hardware also includes an address translation mechanism 175.
  • a write pipe 176 temporarily stores data during one write cycle for application of that data to the system bus 30 during the next cycle to speed up operation of the System/88.
  • System/88 bus logic 177 of the type described in US patent number 4,453,215 couples the translation unit 175 and the write pipe 176 to the system bus 30 in a manner described generally in the above mentioned patent.
  • a similar System/88 bus logic unit 178 couples the storage control interface 155 to the system bus 30.
  • a buffer 180, a programmable read only memory 181, a store 182 and a register set 183 are coupled to the PE62 for use during initialization of the System/88 and the System 370.
  • PROM 181 has system test code and IDCODE required to boot the system from a power on sequence.
  • PROM 181 has the synchronization code for S/88.
  • Register 183 has the system status and control register.
  • Two of the S/370 chip sets are mounted on the same physical board, brought into synchronization, and execute programs in lock-step, to provide board self checking.
  • the STC Bus 157 and a channel 0, 1 bus will be monitored for potential failures so the S/370 processor cannot propagate an error to another field replaceable unit.
  • the BCU 156 and adapter 154 of interface 89 allow each processor (PE62, PE85) to have appropriate control over the other processor so that neither operating system is in full control of the system.
  • Each processor's functions are in part controlled by the interface 89 and microcode running in each processor.
  • the adapter 154 (Fig. 13) interfaces the S/370 processor 85 to the BCU 156 via its output Channels 0, 1.
  • the Channels include a pair of asynchronous two-byte-wide data buses 250, 251.
  • the buses 250, 251 are coupled to the synchronous four-byte-wide data path in processor bus 170 via a pair of 64 byte buffers 259, 260.
  • Data is transferred from the BCU 156 to adapter 154 (and S/370 main storage 162) via bus 251 and from the adapter 154 to the BCU 156 via bus 250.
  • the adapter 154 includes the following registers:
  • the adapter 154 is the interface between PE85 and the BCU 156. Logically, adapter 154 provides the following services to the BCU 156:
  • the BCU 156 has access to the complete storage 162, including its IOA area 187 (Fig. 28).
  • Adapter 154 performs address boundary checking (ACB check) between the IOA area 187 and the user area 165 while key checking is done by cache controller 153 after receiving key, command and storage 162 address data via the processor bus 170 from adapter 154. If the addressed line of data to be stored is held in the cache, then data is stored in the cache. Otherwise controller 153 transfers the data to main store 162. For data fetches the same mechanism applies in cache controller 153.
  • 1/0 command and message transfers between PE85 and BCU 156 are done through predefined storage 162 locations (mailbox area 188 and inbound message queue 189) shown in Fig. 28.
  • the BCU 156 fetches I/O commands from the mailbox area 188 of 16 bytes.
  • the address for accesses to the mailbox area is computed as follows:
  • the first two terms are supplied by base register 110 of adapter 154, the last by the BCU 156.
  • the queue length is set by two bits in the base register 110 to 1, 2, 4 or 8kB (i.e. 64 to 512 entries). Its base is set in the base register 110 to a boundary of two times the buffer size (i.e. 2-16 kB respectively).
  • the inbound message queue 189 stores all messages received via the BCU 154 in chronological order. Each entry is 16 bytes long.
  • the read pointer (RPNTR) and write pointer (WPNTR) in registers 111, 112 are used by the BCU 156 for reading entries from and writing entries into the queue 189.
  • the PE85 accesses the readpointer by a sense-operation.
  • the base address in register 110 plus WPNTR points to the next queue-entry to be written and base address plus RPNTR points to the next queue-entry to be read.
  • the validity of data stored in the mailbox area 188 is signaled from the PE 85 to the BCU 156 and vice versa by the following mechanisms:
  • the BCU 156 makes a request when a problem occurs either during execution of an order initiated by the PE 85 or at any other time. It causes an exception in the PE 85, if not selectively masked.
  • Adapter 154 matches the transfer speed of the asynchronous adapter channels 0,1 to the synchronous processor bus 170. Therefore the BCU 156 is supported by 64 byte data buffers 259, 260 in adapter 154 for data transfer to and from BCU 156 respectively.
  • the array has a 4-byte port to the channel 0,1 bus and to the processor bus 170.
  • Synchronous registers 113, 114 buffer data transferred between BCU 156 and the buffer arrays 260, 259.
  • Bus receive and send registers 115 and 116 store data received from and transferred to processor bus 170 respectively.
  • a store operation (1/0 Data Store, Queue Op) is started by the BCU 156 sending to the adapter 154 the command/byte count, protection key and storage address via the channel 1 bus.
  • the command/byte count is received on the command-bus 252 (Fig. 13) and stored into the command register 125.
  • Key and address data are received from BCU 156 via the address/data-bus 251 (Fig. 13) and stored into the key/addr-register 123.
  • the array write and read address pointers are set to their starting values in register 128.
  • the number of data transfers (2 bytes at a time) on the bus 251 are determined by the byte count. With one store operation, up to 64 bytes of data can be transferred.
  • the storage address of any byte within a store operation may not cross a 64 byte address boundary.
  • the command/address is followed by data cycles on the bus 251. All data is collected in the 64 byte buffer 260. After the last data is received from the BCU 156, the adapter 154 performs first an internal priority check (not shown) for the two data buffers 259, 260 and then requests mastership (not shown) on the processor bus 170, where adapter 154 has the highest request priority.
  • both buffers 259, 260 request a transfer at the same time the internal priority control grants the bus 170 first to buffer 259 and then without an arbitration cycle to buffer 260, i.e.: reads have priority over writes.
  • command/byte count, protection key and the starting address are transferred to cache controller 153.
  • the command transfer cycle is followed by data transfer cycles.
  • Cache controller 153 performs the protection key checking. A key violation will be reported to adapter 154 in the bus 170 status. Other check conditions detected by cache controller 153 and main store 162 are reported as ANY-CHECK status. A key violation and status conditions detected by adapter 154 will be sent to the BCU 156 in a status transfer cycle.
  • Each main store address received from the BCU 156 is compared with the address kept in the ACB register to determine whether the access is to the IOA 187 or customer area 165 of storage 162.
  • a "customer" bit received along with each command from the BCU 156 determines whether the main storage access is intended for the IOA area 187 or customer area 165 and checks for improper accesses.
  • BNA Buffer Not Available
  • Read operations (1/0 Read, Mailbox Read) are started by the BCU 156 in a manner essentially the same as store operations.
  • the adapter 154 internal priority check is performed and processor bus 170 mastership is requested. If bus mastership is granted, command/byte count, protection key and the main store starting address is transferred to cache controller 153 to initiate the read cycle.
  • Adapter 154 loads the requested data first in its buffer 259 and then, on BCU request via the bus 250, to the BCU 156. Status is reported with each data transfer.
  • the status conditions and reporting mechanism for store operations apply to read operations.
  • PE85 can access most of the registers in adapter 154 with both sense (read) and control (write) operations via the bus 170.
  • the command is transferred to adapter 154 and latched into the register 129.
  • the sense multiplexor 126 is selected according to the command; and the command is loaded into the BSR 116 to have the expected data valid in the following bus 170 cycle.
  • adapter 154 sends data with good parity back to the PE85, but raises a check condition on the Key/Status bus. This function can be tested with a specific sense codepoint.
  • the BUS 170 command will be followed by data, which is loaded into the target register in the next cycle.
  • adapter 154 forces a clock stop.
  • the base register 110 contains the base-address used for queue and mailbox addressing and the queue length code.
  • the queue starts at the base address, the mailbox-area at base + queue length.
  • the RPNTR and WPNTR registers 111 and 112 registers contain the offset from the base address to the next queue entry to be accessed for a read or write respectively.
  • the read pointer and write pointer are concatenated with the base-address by sense multiplexer 126 in adapter 154. Therefore the word returned by the sensed operation is the complete address of the next queue-entry to be accessed.
  • the I/O Status Register contains the following bits (in addition to others, not described herein):
  • the customer access bit which is part of the command/address received from the BCU 156, determines if the storage access will be in the IOA or customer storage area. If the customer access bit is '0', the page address for the storage access must be within the IOA area 187. No Key checking will be done for these accesses, hence the adapter hardware forces the Key to zero (matches with all key entries).
  • the page address for the storage access must be within the customer storage area 165. Otherwise an ACB check condition is raised for the access.
  • the PE85 uses Message Commands to read (sense) or write (control) the adapter 154 registers.
  • the DST field for the PU-BCU Interface is X'8'.
  • Adapter 154 will not decode the SRC and MSG field since there is no information contained for command execution.
  • the Reg1 and Reg2 bits will define respectively the register in adapter 154 to be written into and read from.
  • the adapter channel 0 and adapter channel 1 are high speed interconnections from the I/O adapter 154 to the bus control unit 156.
  • Channel 0 includes:
  • Channel 1 includes an address/data bus 251, a command/status bus 252 and tag up and tag down lines 262c and 262d.
  • Channel 0 is used for data transfers from S/370 storage 162 (and PE 85) to BCU 156 and Channel 1 is used for data transfers from BCU 156 to storage 162 (and PE 85).
  • the channel buses 249, 250, 251 and 252 originate in the 1/0 adapter 154 which is essentially a pair of data buffers with control logic capable of storing up to 64 bytes of data each.
  • the buses terminate in the BCU 156.
  • the I/O adapter 154 serves as speed match between the S/370 internal processor bus 170 with its full-word format (32 bits) and the slower buses 249-252 with their half word format (16 bits).
  • Each channel is organized in two portions, the two-byte wide (half-word) data bus (250, 251) and the half-byte wide (4-bit) command/status bus (249, 252).
  • Tag signals provide the means to control the operations via request/response, and special signals.
  • the data transfer over each channel occurs always in two cycles (to transfer four bytes over the two-byte bus).
  • all data transfer is between S/370 main storage 162 and the I/O subsystem including BCU 156.
  • the BCU 156 is the master, that is, it initiates all transfer operations once the PE 85 has signaled the need for it.
  • the command/status bus (249, 252) is used during a select cycle to define the transfer direction (fetch/store), and the amount of data to be transferred.
  • the address/data bus (250, 251) serves to transfer the main storage address during the select cycle and delivers data during the actual transfer cycle. It is also used to indicate specific areas 188, 189 in storage 162 known as "mailbox" and "message queue”. These areas allow the PE 85 to exchange certain information with the BCU 156.
  • the status is transferred over the command/status bus 249 together with the first two bytes of data on bus 250. This status indicates any address check, key check, etc, or is zero to indicate a successful operation.
  • Figs. 14A and 14B show the logical usage of the bus portions during subcycle 1 and subcycle 2 of fetch and store operations respectively, wherein:
  • tag lines are used for data transfer operations:
  • the PE 85 When the PE 85 detects a Start 1/0 instruction (SIO) in the instruction stream, it alerts the I/O subsystem, i.e. BCU 156, about the need for an 1/0 operation by activating the "PU to BCU Request" line 256a.
  • This tag causes the BCU 156 to look into the "mailbox" 188 within store 162 to find out whether this operation is a fetch or a store, how many bytes are to be transferred, etc.
  • the mailbox actually contains the channel SIO, CUA, CAW and command word (CCW) of the pertinent 1/0 operation.
  • Store operations are generally those where the BCU 156 sends data to the PE 85.
  • This "data" is either the command/key/address which is sent in the select cycle or the "real" I/O data to be stored in main storage 162. In both cases, the sequence of events is the same.
  • Figs. 15A-C diagrammatically illustrate in a generalized form, for the following description, the manner in which data and status information are gated in and out of thirty-two bit buffers/registers in adapter 154 and BCU 156 and in which the higher order (left) and lower order (right) bits of the information are placed on the eighteen bit channel 0, 1 buses of the adapter 154.
  • Figs. 25 and 26 provide a specific set of signals for data transfers between BCU 156 and adapter 154.
  • the BCU 156 places the data for the first cycle onto the bus 251. If this is a select cycle for a main storage data operation, a command, a byte count, an access key and the first byte of the main storage address is placed on the command/status bus 252 and the address/data bus 251, respectively. If this is the select cycle for a mailbox lookup, no main storage address is placed since the command indicates the mailbox which is in a fixed location. The first subcycle is maintained valid on the bus for two subcycle times.
  • the BCU 156 raises the "Tag Up" signal line.
  • the Tag Up line 262a causes the adapter 154 to store the first two bytes in the left half of register 113.
  • the BCU 156 places the data (second two bytes) for the next subcycle on the address/data bus 251 for storage in the other half of the register 113 adapter 154. This data is either the remainder of a main storage address, or an offset (if the shot belongs to a mailbox lookup select cycle).
  • the BCU 156 holds the second two bytes for three BCU clock cycles, then drops the "Tag Up" signal.
  • Fetch operations are generally those where the BCU 156 demands data from the main storage data space 162, from the microcode area in main storage 162, or from the mailbox or the message queue.
  • a select cycle must precede such a fetch operation to instruct the logic of adapter 154 about the operation it must execute.
  • the select cycle is performed by placing command/key/address on the bus 249 in a manner similar to the store operation using bus 252, except that the command on the command/status bus 249 is a "fetch" command.
  • the BCU 156 With the beginning of the next clock cycle (after completion of the select cycle) the BCU 156 raises the "Tag Up" signal and maintains it for three BCU clock cycles (Fig. 15B). Tag up demands data from the buffer. Data will be available one cycle later if the buffer can deliver data. Since the operation is semi- synchronous, the BCU 156 assumes that the first two bytes of data are maintained valid on the bus for two cycles, then there is a switch-over time of one cycle, and thereafter the second two bytes of data can be gated to the BCU 156.
  • the adapter 154 maintains "Tag Down” until the first data word (four bytes) is available. At that instant, the adapter 154 places the first two bytes onto the bus 250 and drops “Tag Down". The falling edge of the "Tag Down" signal triggers the BCU's logic 253.
  • the BCU 156 assumes that the first bytes are valid for two cycles following the dropping of "Tag Down," and thereafter the second two bytes are available. Depending on the count that is set up during the select cycle up to 60 bytes can follow, two bytes at a time.
  • the BCU 156 raises the "BCU to PU Acknowledge" signal on line 256b to the adapter 154 to reset the PU to BCU request on line 256a that started the operation.
  • the inbound message queue 189 stores all messages sent by the BCU in chronological order.
  • the Bus Control Unit (BCU) 156 is the primary coupling hardware between the S/370 processor 85 and its associated S/88 processor 62 which is utilized to perform the S/370 1/0 operations.
  • the BCU 156 includes means which interacts with an application program (EXEC370) and microcode (ETIO) running on the S/88 processor 62 to present interrupts to the processor 62 and to asynchronously uncouple the processor 62 from its associated hardware and to couple the processor 62 to the BCU 156, all transparent to the S/88 operating system.
  • the transparent interrupt and uncoupling functions are utilized to permit the direct coupling of the S/370 and S/88 processors for the efficient transfer of S/370 I/O commands and data from the S/370 processor 85 to the S/88 processor 62 for the conversion of the commands and data to a form usable by the S/88 processor 62 to perform the desired S/370 I/O operations.
  • EXEC370 and ETIO may both be either microcode or application program and stored in either store 174 or cache 173.
  • the BCU 156 includes bus control unit interface logic and registers 205, a direct memory access controller (DMAC) 209 and a local store 210.
  • Local address and data buses 247, 223 couple store 210 to the PE62 address, data buses 161 A, 1610 via driver/receiver circuits 217, 218 and to the interface logic 205.
  • DMAC 209 is coupled to address bus 247 via latches 233 and to data bus 223 via driver/receivers 234.
  • DMAC 209 in the preferred embodiment is a 68450 DMA controller described in greater detail below.
  • DMAC 209 has four channels 0-3 which are coupled to the interface logic 205 (Fig. 17) by respective Request and Acknowledge paths, each dedicated to a specific function;
  • Channel 0 transfers S/370 I/O commands from a mail box area 188 (Fig. 28) in S/370 storage 162 to local store 210 (MAILBOX READ).
  • Channel 1 transfers S/370 data from storage 162 to store 210 (S370 1/0 WRITE).
  • Channel 2 transfers data from store 210 to storage 162 (S/370 1/0 Read).
  • Channel 3 transfers high priority S / 88 messages from Store 210 to a message queue area 189 (Fig. 28) in Storage 162 (Q Message WRITE).
  • the bus adapter 154 has two channels 0 and 1.
  • Adapter channel 0 handles the MAILBOX READ and S/370 1/0 WRITE functions of DMAC channels 0, 1 (i.e., data flow from S/370 to BCU 156).
  • Adapter channel 1 handles the S/370 1/0 READ and Q MESSAGE WRITE functions of DMAC channels 2, 3 (i.e., data flow from BCU 156 to S/370).
  • the DMAC 209 is preferably of the type described (MC68450) in the M68000 Family Reference Manual, FR68K/D, Copyright Motorola, Inc., 1988.
  • the DMAC 209 is designed to complement the performance and architectural capabilities of Motorola M68000 Family microprocessors (such as the M68020 processor element 62 of the present application) by moving blocks of data in a quick, efficient manner with minimum intervention from a processor.
  • the DMAC 209 performs memory-to-memory, memory-to-device, and device-to-memory data transfers.
  • DMAC direct memory access
  • DMA direct memory access
  • the main purpose of a DMAC such as 209 in any system is to transfer data at very high rates, usually much faster than a microprocessor under software control can handle.
  • DMA direct memory access
  • the term direct memory access (DMA) is used to refer to the ability of a peripheral device to access memory in a system in the same manner as a microprocessor does.
  • the memory in the present application is local store 210. DNA operation can occur concurrently with other operations the the system processor needs to perform, thus greatly boosting overall system performance.
  • the DMAC 209 moves blocks of data at rates approaching the limits of the local bus 223.
  • a block of data consists of a sequence of byte, word, or long-word operands starting at a specific address in storage with the length of the block determined by a transfer count.
  • a single channel operation may involve the transfer of several blocks of data to or from the store 210.
  • any operation involving the DMAC 209 will follow the same basic steps: channel initialization by PE62. data transfer, and block termination.
  • the processor PE62 loads the registers of the DMAC with control information, address pointers, and transfer counts and then starts the channel.
  • the DMAC 209 accepts requests for operand transfers and provides addressing and bus control for the transfers.
  • the termination phase occurs after the operation is complete, when the DMAC indicates the status of the operation in the status register CSR.
  • the DMAC 209 will be in one of three operating modes:
  • the DMAC can perform implicit address or explicit address data transfers.
  • explicit transfers data is transferred from a source to an internal DMAC holding register, and then on the next bus cycle it is moved from the holding register to the destination.
  • Implicit transfers require only one bus cycle because data is transferred directly from the source to the destination without internal DMAC buffering.
  • the memory address and device address registers MAR and DAR are initialized by the user to specify the source and destination of the transfer. Also initialized is the memory transfer count register to count the number of operands transferred in a block.
  • the two chaining modes are array chaining and linked array chaining.
  • the array chaining mode operates from a contiguous array in store 210 consisting of memory addreses and transfer counts.
  • the base address register BAR and base transfer count register BTC are initialized to point to the beginning address of the array and the number of array entries, respectively.
  • the base transfer count is decremented and the base address is incremented to point to the next array entry.
  • the base transfer count reaches zero, the entry just fetched is the last block transfer defined in the array.
  • the linked array chaining mode is similar to the array chaining mode, except that each entry in the memory array also contains a link address which points to the next entry in the array. This allows a non- contiguous memory array. The last entry contains a link address set to zero.
  • the base transfer count register BTC is not needed in this mode.
  • the base address register BAR is initialized to the address of the first entry in the array.
  • the link address is used to update the base address register at the beginning of each block transfer.
  • This chaining mode allows array entries to be easily moved or inserted without having to reorganize the array into sequential order. Also, the number of entries in the array need not be specified to the DMAC 209. This mode of addressing is used by DMAC 209 in the present application for accessing free work queue blocks (WQB) from a link list in a manner described in detail below.
  • WQB free work queue blocks
  • the DMAC 209 will interrupt the PE62 for a number of event occurrences such as the completion of a DMA operation, or at the request of a device using a PCL line 57a-d.
  • the DMAC 209 holds interrupt vectors in eight on-chip vector registers for use in the PE62 vectored interrupt structure. Two vector registers, normal interrupt vector (NIV) and error interrupt vector (EIV), are available for each channel.
  • NUV normal interrupt vector
  • EIV error interrupt vector
  • Each channel is given a priority level of 0, 1, 2, or 3, i.e., channel 0, 1, 2, 3 are assigned priority levels 0, 2, 2, 1 respectively (priority level 0 is highest).
  • Requests are externally generated by a device or internally generated by the auto-request mechanism of the DMAC 209. Auto-requests may be generated either at the maximum rate, where the channel always has a request pending, or at a limited rate determined by selecting a portion of the bus bandwidth to be available for DMA activity. External requests can be either burst requests or cycle steal requests that are generated by the request signal associated with each channel.
  • the DMAC 209 contains 17 registers (Fig. 18) for each of the four channels plus one general control register GCR, all of which are under software control.
  • the DMAC 209 registers contain information about the data transfers such as the source and destination address and function codes, transfer count, operand size, device port size, channel priority, continuation address and transfer count, and the function of the peripheral control line.
  • One register CSR also provides status and error information on channel activity, peripheral inputs, and various events which may have occurred during a DMA transfer.
  • the feral control register GCR selects the bus utilization factor to be used in limited rate auto-request DMA operations.
  • the input and output signals are functionally organized into the groups as described below (Ref. Fig. 19A).
  • the address/data bus (A8-A23, DO-D15) 248 a 16-bit bus, is time multiplexed to provide address outputs during the DNA mode of operation and is used as a bidirectional data bus to input data from an external device (during a PE62 write or DMAC read) or to output data to an external device (during an PE62 read or a DMAC write).
  • This is a three-state bus and is demultiplexed using external latches and buffers 233, 234 controlled by the multiplex control lines OWN and DDIR.
  • Lower address bus lines A1 through A7 of bus 247 are bidirectional three-state lines and are used to address the DMAC internal registers in the MPU mode and to provide the lower seven address outputs in the DNA mode.
  • Function code lines FCO through FC2 are three-state output lines and are used in the DMA mode to further qualify the value on the address bus 247 to provide separate address spaces that may be defined by the user.
  • the value placed on these lines is taken from one of the internal function code registers MFC, DFC, BFC, depending on the register that provides the address used during a DMA bus cycle.
  • Asynchronous bus control lines control asynchronous data transfers using the following control signals: select address strobe, read/write, upper and lower data strobes, and data transfer acknowledge. These signals are described in the following paragraphs.
  • SELECT input line 296 is used to select the DMAC 209 for an MPU bus cycle. When it is asserted, the address on A1-A7 and the data strobes (or AO when using an 8-bit bus) select the internal DMAC register that will be involved in the transfer. SELECT should be generated by qualifying an address decode signal with the address and data strobes.
  • ADDRESS STROBE (AS) on line 270b is a bidirectional signal used as an output in the DMA mode to indicate that a valid address is present on the address bus 161. In the MPU or IDLE modes, it is used as an input to determine when the DMAC can take control of the bus (if the DMAC has requested and been granted use of the bus).
  • READ/WRITE is a bidirectional signal (not shown) used to indicate the direction of a data transfer during a bus cycle.
  • a high level indicates that a transfer is from the DMAC 209 to the data bus 223 and a low level indicates a transfer from the data bus to the DNAC 209.
  • a high level indicates a transfer from the addressed memory 210 to the data bus 223 and a low level indicates a transfer from the data bus 223 to the addressed memory 210.
  • UPPER AND LOWER DATA STROBE bidirectional lines (not shown) indicate when data is valid on the bus and what portions of the bus should be involved in a transfer D8-15 or DO-7.
  • DATA TRANSFER ACKNOWLEDGE (DTACK) bidirectional line 265 is used to signal that an asynchronous bus cycle may be terminated.
  • this output indicates that the DMAC 209 has accepted data from the PE62 or placed data on the bus for PE62.
  • this input 265 is monitored by the DMAC to determine when to terminate a bus cycle. As long as DTACK 265 remains negated, the DMAC will insert wait cycles into a bus cycle and when DTACK 265 is asserted, the bus cycle will be terminated (except when PCL 257 is used as a ready signal, in which case both signals must be asserted before the cycle is terminated).
  • Multiplex control signals on lines OWN and DDIR are used to control external multiplex/demultiplex devices 233, 234 to separate the address and data information on bus 248 and to transfer data between the upper and lower halves of the data bus 223 during certain DMAC bus cycles.
  • OWN line is an output which indicates that the DMAC 209 is controlling the bus. It is used as the enable signal to turn on the external address drivers and control signal buffers.
  • BUS REQUEST (BR) line 269 is an output asserted by the DMAC to request control of the local bus 223, 247.
  • BUS GRANT (BG) line 268 is an input asserted by an external bus arbiter 16 to inform the DMAC 209 that it may assume bus mastership as soon as the current bus cycle is completed.
  • the two interrupt control signals IRQ and IACK on lines 258a and 258b form an interrupt request/acknowledge handshake sequence with PE62 via interrupt logic 212.
  • INTERRUPT REQUEST (IRQ) on line 258a is an output is asserted by the DNAC 209 to request service from PE62.
  • INTERRUPT ACKNOWLEDGE (IACK) on line 258b is asserted by PE62 via logic 216 to acknowledge that it has received an interrupt from the DMAC 209.
  • IACK INTERRUPT ACKNOWLEDGE
  • the DMAC 209 will place a vector on DO-D7 of bus 223 that will be used by the PE 62 to fetch the address of the proper DMAC interrupt handler routine.
  • the device control lines perform the interface between the DMAC 209 and devices coupled to the four DMAC channels. Four sets of three lines are dedicated to a single DMAC channel and its associated peripheral; the remaining lines are global signals shared by all channels.
  • REQUEST REQO THROUGH REQ3 inputs on lines 263a-d are asserted by logic 253 to request an operand transfer between main store 167 and store 210.
  • ACKNOWLEDGE (ACKO THROUGH ACK3) outputs on lines 264a-d are asserted by the DMAC 209 to signal that an operand is being transferred in response to a previous transfer request.
  • PERIPHERAL CONTROL LINES (PCLO THROUGH PCL3) 257a-d inclusive are bidirectional lines between interface logic 253 and DMAC 209 which are set to function as ready, abort, reload, status, interrupt, or enable clock inputs or as start pulse outputs.
  • DATA TRANSFER COMPLETE (DTC) line 267 is an output asserted by the DMAC 209 during any DMAC bus cycle to indicate that data has been successfully transferred.
  • DONE DONE
  • This bidirectional signal is asserted by the DMAC 209 or a peripheral device during DNA bus cycle to indicate that the data being transferred is the last item in a block.
  • the DMAC will assert this signal during a bus cycle when the memory transfer count register is decremented to zero.
  • the BCU interface logic 205 (Fig. 16) has been separated into various functional units for ease of illustration and description in Figs. 19A-C.
  • the logic 205 includes a plurality of interface registers interposed between the local data bus 223 and the adapter channels 0, 1 for increasing the speed and performance of data transfers between the adapter 154 and the BCU 156.
  • the hardware logic 253 of interface 205 together with DMAC 209, the address decode and arbitration logic 216 and address strobe logic 215 control the operations of the BCU 156.
  • the interface registers include a channel 0 read status register 229 and a channel 1 write status register 230 coupled to the channel 0 and 1 command status buses 249, 252 for holding the status of data transfers between adapter 154 and BCU 156.
  • 1 address/data registers 219,227 hold the S/370 address for transfer to adapter 154 during S/370 I/O data transfers.
  • Register 227 also holds succeeding I/O data words (up to 4 bytes) of data transfers (up to 64 bytes per address transfer) to adapter 154 after each address transfer.
  • Channel 0 read buffer receives I/O data transferred from adapter 154 during BCU mailbox read and S/370 I/O write operations.
  • BSM read/write select up byte counters 220, 222 and BSM read/write boundary counters 221, 224 hold byte counts for transfer of data from the BCU 156 to adapter 154. Both counters are required for each channel to avoid the crossing of S/370 sixty-four byte address boundaries by data transfers.
  • counters 220, 222 initially store the total byte count to be transferred for an I/O operation (up to 4KB) and are used to transfer count values to registers 214, 225 to partially form a S/370 starting address only for the last block (64 bytes) transfer, i.e. the last command/data transfer operation.
  • the boundary counters 221, 224 are used to present (in part) a starting S/370 address whenever a boundary crossing is detected by the BCU 156 for any single command data transfer operation or when the byte count is greater than 64 bytes.
  • the counters 220, 221, 222 and 224 are appropriately decremented after each data transfer over channel 0 or 1.
  • a queue counter 254 provides a similar function for message transfers (up to sixteen bytes) to S/370 storage via adapter 154.
  • the addresses for selecting the above interface registers are in the store 210 address space, Fig. 23C, and are selected by decoding the address on bus 247 in a well known manner.
  • a signal on PU to BCU request line 256a from the adapter 154 to logic 253 notifies BCU 156 that a S/370 mailbox read request is ready. This signal is not reset by a BCU PU acknowledge signal on line 256b until the mailbox information has been stored into local store 210.
  • Tag up and tag down lines 262a-d are used for strobing data between the BCU 156 and adapter 154 over adapter channels 0, 1.
  • Handshake signals are provided between the BCU logic 253 and DMAC 209.
  • BCU logic makes service requests on lines 263a-d, one for each DMAC channel.
  • DMAC responds with acknowledge signals on lines 264a-d.
  • Other lines such as select 270, data transfer acknowledge 265, peripheral control lines 257a-d, data transfer complete 267 have been described above with respect to DMAC 209.
  • the "uncoupling" logic decodes the virtual address applied to the S/88 processor address bus 161 A during each instruction execution cycle. If one of the block of preselected S/88 virtual addresses assigned to the BCU 156 and its store 210 are detected, the address strobe (AS) signal from the S/88 processor 62 is gated to the BCU 156 rather than to the associated S/88 hardware. This action prevents the S/88 Operating System and hardware from knowing a machine cycle has taken place, that is the action is transparent to the S/88.
  • AS address strobe
  • the S/88 processor 62 is coupled to control the BCU 156 during this machine cycle, the AS signal and the preselected address being used to select and control various components in the BCU 156 to perform a function, related to S/370 1/0 operations.
  • Special application code (EXEC370) running on the S/88 processor 62 initiates communication with the S/370 processor 85 by placing these preselected virtual addresses on the S/88 bus 161A to direct the BCU 156 to perform operations to effectuate said communication.
  • the DMAC 209 and other logic in the BCU 156 present interrupts to the S/88 at a specified level (6) calling this special application code into action as required.
  • the presentation of each interrupt is transparent to the S/88 Operating System.
  • one partner unit is a connected sandwich of a modified dual S/88 processor board with a dual S/370 processor board containing dual local stores, DMACs, and custom logic.
  • the like elements of this dual sandwiched board operate in parallel, in full synchronism (lock-step) for fault-detection reasons.
  • This entire sandwich normally has an identical partner sandwich, and the partners run in lock-step, thus appearing as a single fault-tolerant entity. It is sufficient to the following discussion to consider this doubly- replicated hardware as a single operational unit as shown in Fig. 21.
  • up to eight of these operational units 295 to 295-8 may reside within a single module enclosure, sharing main memory, I/O facilities, and power supplies, under the control of a single copy of the S/88 Operating System.
  • the unit 295 (and each other unit 295-2 and 295-8) corresponds to a pair of partner boards such as boards 21, 23 of Fig. 7.
  • the S/88 processor units 62 to 62-8 operate as multi-processors sharing the S/88 workload, but the S/370 units 85 to 85-8 operate separately and independently and do not intercommunicate.
  • Each S/370 unit runs under control of its own Operating System, and has no 'knowledge' of any other CPU in the enclosure (either S/370 or S/88).
  • each interrupt (from I/O, timers, program traps, etc.) is presented on the common bus 30 to all S/88 processor units in parallel; one unit accepts the responsibility for servicing it, and causes the other units to ignore it.
  • the servicing CPU unit there is a single vector table, a single entry point (per vector) within the Operating System for the handler code, and disposition of the interrupt is decided and handled by the (single) Operating System.
  • a requirement is that a DMAC interrupt must be handled only by the S/88 processor 62 to which that DNAC, BCU, and S/370 is attached, so that the multiple S/370 units 85 to 85-8 cannot interfere with each other.
  • the DNAC IRQ line 258a is wired directly to the S/88 processor 62 to which the DMAC 209 is attached and does not appear on the common S/88 bus 30, as do all of the normal S/88 interrupt request lines.
  • a given S/88 processor 62 is dedicated to the S/370 to which it is directly attached.
  • Eight user vector locations within the main S/88 vector table are reserved for use by the DMACs. and these vectors are hard-coded addresses of eight DMAC interrupt handlers which are added to the S 88 Operating System. These eight interrupt handlers are used by all S/88 processors to process interrupts presented by all DMACs for the associated S/370 processors.
  • Each DMAC such as 209 has a single interrupt request (IRQ) output signal and eight internal vector registers (two per channel, one each for normal operations and DMAC-detected errors). At initialization time (described later), these DMAC vector register values are programmed to correspond to the eight reserved main vector-table locations mentioned above. Thus a DMAC may request one of eight handler routines when it presents IRQ. These handlers access the DMAC BCU hardware, queues, linked lists, and all control parameters by presenting virtual addresses that lie within the address range of the 'hidden' local store 210.
  • IRQ interrupt request
  • the hardware design ensures that each S/88 processor such as 62 can access its own store such as 210 and no others, even though a common virtual-address uncoupling 'window' is shared among multiple S/370 units. That is, the S/88 virtual address space 007EXXXX is used by all S/88-S/370 multiprocessors in a module even though each partnered unit such as 21, 23 has its dedicated S/88 physical storage as shown in Fig. 10.
  • all of the DMACs 209 to 209-8 are programmed identically as regards these eight vector registers, and all share the eight reserved vectors in the main vector table, as well as the handler routines. Differentiation, as well as uncoupling, occur at each access to the store such as 210.
  • the complete interrupt design thus accomplishes intermittent 'dedicated upon demand' servicing of the S/370 DMAC interrupts, with isolation and protection for multiple S/370 units, by usurping individual processor facilities from a multiprocessing system environment which uses a different interrupt servicing philosophy, with essentially no impact upon the multiprocessing system operation and no significant changes to the multiprocessing Operating System.
  • each DMAC interrupt mechanism For a more detailed operation of each DMAC interrupt mechanism, attention is directed to Figs. 19A and 20.
  • a peripheral device such as DMAC 209 having selection vectors presents an interrupt request (TRQ) to the S/88 processor 62
  • TRQ interrupt request
  • a single IRQ line 258a is made active by the device.
  • This IRQ line is wired to an encoding circuit 293 in a manner specified by the S/88 processor architecture, so as to present an encoded interrupt request to the S/88 processor 62 via input pins IPLO-IPL2 at a specific priority level 6.
  • the processor 62 effectively decides when it can service the interrupt, using priority masking bits kept in the internal status register. When ready, the processor 62 begins a special 'Interrupt Acknowledge' (IACK) cycle.
  • IACK Interrupt Acknowledge'
  • a unique address configuration is presented on the address bus 161A in order to identify the type of cycle and priority level being serviced. This is also effectively a band for a vector number from the interrupting device. All requesting devices compare the priority level being serviced with their own, and the device with a matching priority gates a one-byte vector number to the data bus 161 D for the processor 62 to read.
  • the processor 62 saves basic internal status on a supervisor stack and then generates the address of the exception vector to be used. This is done by internally multiplying the device's vector number by four, and adding this result to the contents of the internal Vector Base Register, giving the memory address of the exception vector. This vector is the new program counter value for the interrupt handler code.
  • the first instruction is fetched using this new program counter value, and normal instruction decoding and execution is resumed, in supervisor state, with the processor 62 status register set to the now-current priority level.
  • the DMAC 209 interrupts in the preferred embodiment are wired to priority level six, and conform entirely to the processor 62 architecture.
  • the DMAC 209 has eight vector numbers programmed internally, and eight separate handler routines are used.
  • the decode and arbitration logic 216 (Fig. 19A) and AS control logic 215 control this interrupt function during the IACK cycle in addition to providing the S/88 processor 62 uncoupling function.
  • Fig. 20 shows details of logic 215 and 216 of Fig. 19A.
  • the address strobe line 270 from PE62 is coupled to one input of control logic 15.
  • Logic 216 has a pair of decode circuits 280, 281.
  • the output 282 of circuit 280 is coupled to logic 215; the output 283 of circuit 281 is also coupled to logic 215 via AND gate 291 and line 287.
  • decode circuits 280, 281 permit the address strobe signal (AS) on line 270 to pass through logic 215 to line 270a which is the normal address strobe to S/88 hardware associated with PE62.
  • AS address strobe signal
  • the decode logic 280 puts a signal on line 282 to block the AS signal on line 270a and sends AS to the BCU 156 via line 270b.
  • the decode logic 280 may also be designed to detect an appropriate function code on lines FCO-2; however this is merely a design choice. Figs.
  • the blocking signal on line 282 is applied to OR circuit 284 to produce a PE62 local bus request signal on line 190 to the arbitration logic 285.
  • Logic 285 will grant the request to PE62 only if DMAC 209 has not already placed a request on line 269.
  • the PE62 bus grant line 191 is activated if there is no DMAC request.
  • the PE62 bus grant signal on line 191 raises ENABLE lines 286a, b (Fig. 19A) via logic 253 to couple PE62 buses 161A, D to local buses 247, 223 via drivers 217 and driver/receivers 218 in preparation fora PE62 operation with BCU 156.
  • Data and Commands may be transferred between the PE62 and elements of the BCU while the processor buses 161A, D are coupled to the local buses 247, 223 under control of the instruction being executed by PE62.
  • the application program EXEC370 and the ETIO firmware contain such instructions.
  • logic 285 gives the DMAC 209 priority over the PE62 request on line 190; the DMAC bus grant signal on line 268 is returned to DMAC 209; and the local bus 247, 223 is connected between either the local store 210 and adapter channels 0, 1 via the high speed interface registers or between the DMAC 209 and the local store 210 in preparation for a DMAC operation with BCU 156.
  • logic 215, 216 uncouples the S/88 processor 62 from the associated hardware (e.g., 175, 176, 177) and couples it to the BCU 156 when an address 007EXXXX is decoded by logic 280. This uncoupling is transparent to the S/88 operating system.
  • the decode logic 281 (and associated hardware) blocks address strobe AS from line 270a and initiates a local bus request to the arbitration logic 285 during a DMAC 209 interrupt sequence to PE62.
  • DNAC 209 places an interrupt signal on line 258a, it is applied to PE62 via OR circuits 292a and 292, level 6 input of the S/88 interrupt priority logic 293 and lines IPLO-2.
  • PE62 responds with an interrupt acknowledge cycle.
  • Predetermined logical bits (which include the value of the interrupt level) are placed on output FCO-2 and address bus 161A (bits A1-3, A16-19), which bits are decoded by logic 281 to produce an output on line 283.
  • This output and the interrupt signal on line 258c cause AND gate 291 to apply a signal to line 287 causing logic 215 to apply AS to the BCU logic 253 via line 270b.
  • the signal on line 287 blocks A-from line 270a and places a PE62 bus request on line 190 via OR circuit 284 to arbitration logic 285. Because the address strobe (AS) signal is blocked from going to the S/88 hardware, this interrupt is transparent to the S/88 Operating System.
  • AS address strobe
  • decode logic 281 When the special IACK bits are received on bus 161A and FCO-2 as described above, decode logic 281 produces an output signal on line 283 to block an address strobe signal on line 270a and to place a PE62 request on arbitration logic 285 via OR circuit 284 and line 190. If there is no DMAC request on line 269, the PE62 bus grant signal is raised on line 191 to AND gate 294-1. The AND gate 294 produces an lACK signal on line 258b to DMAC 209. This alerts the DMAC 209 to present its interrupt vector. The DMAC then places the vector on the local bus and raises 'DTACK' on line 265 to logic 253.
  • Logic 253 in response to the AS signal on line 270b, raises ENABLE signals on lines 286a, 286b to couple the processor buses 161A and D to local buses 248 and 223 via circuits 217, 218 to read the appropriate vector from DMAC 209 into PE62.
  • the DMAC 209 presents interrupt vectors from the least significant byte of its data bus 248 (Fig. 19A) to the S/88 processor data bus 161D, bits 23-16, via driver receiver 234 and bits 23-16 of the local data bus 223.
  • the vector number issued by DMAC 209 is used by the S/88 processor 62 to jump to one of eight interrupt handlers in the S/88 interface microcode ETIO.
  • DTACK on line 265, and logic 253 activates DSACK 0, 1 on lines 266a, b to terminate the PE62 cycle via a pair of OR circuits 288.
  • Lines 266a, b are ORed with standard S/88 DSACK lines 266 c, d to form the ultimate DSACK inputs 266 e, f to PE62.
  • a pair of AND gates 294-2 and 294-3 (Fig. 20) raise IACK signals on lines 258d, e to initiate the transfer of appropriate vector numbers from the BCU156 to the S/88 processing unit 62 via logic 564. 565 of Fig. 49 and local data bus 223.
  • S/88 level 6 interrupt request could be given priority over a DMAC or BCU interrupt request (when they are concurrent) by a minor change in the logic.
  • the time to recognize Power Faults as secondary interrupt sources is more than adequate.
  • the local storage 210 (Fig. 41 C) is of fixed size and is mapped into the S/88 PE 62 virtual-address space.
  • the local storage 210 is divided into three address ranges to differentiate three purposes:
  • the local storage address space includes:
  • the local address decode and bus arbitration unit 216 detects all addresses within this local storage space.
  • the DMAC 209 may, at the same time, be presenting an address within the area 1 above.
  • the DMAC may NOT address areas 2 or 3 above; this is guaranteed by initialization microcode.
  • the BCU 156 monitors all addresses on the local bus and redirects, via control tags, operations having addresses within ranges 2 and 3 to the proper unit (BCU or DMAC) instead of to the local storage 210.
  • the address area of local storage 210 represented by the ranges 2 and 3 above, while present, is never used for storage therein.
  • a fourth operation type is also handled by the local address decode and bus arbitration unit 215:
  • This special operation is detected by address and function code bits that the S/88 PE 62 presents, with the difference that the (architected special) decode is not an address in the range of the local storage 210.
  • the local bus arbitration unit 216 therefore has a special decoder for this case, and assist logic to signal the DMAC to present its pre-programmed interrupt vector.
  • the operation is otherwise similar to the S/88 processor 62 reading a DMAC register.
  • the address bus 247 is selected by PE 62 when the high order digits decode to hexadecimal (H) 007E.
  • the following data is placed on the local data bus 223 by the S/88 processor 62 for a selected DMAC memory transfer count register, and for the BCU 156 to be used in a subsequent BSM Read/Write Select Command:
  • Bits 31-16 (0000 ogbb bbbb bbbb) the byte transfer count are set into the DMAC memory transfer counter:
  • the byte transfer count, (bits 31-16) are set into the DMAC channel 3 memory transfer count register MTC.
  • the BCU 156 captures the data for a subsequent Q Select Up command as follows:
  • S/88 processor 62 local bus operations include:
  • DMAC 209 local bus operations include:
  • BCU 156 local bus operations include:
  • the BCU 156 logic performs the following:
  • the DSACK lines 266a, b are activated by the BCU logic to close out the cycle:
  • the DMAC Bus Request (BR) line 269 from the DMAC 209 is activated for a DMAC or a Link-List load sequence.
  • the BCU 156 performs the following:
  • the address bit assignments from the S/88 processor 62 to the local storage 210 are as follows: low order bits 0,1 (and SIZO, 1 of PE 62, not shown) determine the number and bus alignment of bytes (1-4) to be transferred. Bits 2-15 inclusive are the address bits for storage space 210.
  • the DMAC address bit A2 is used as the low order address bit (double word boundary) to the local storage 210. Since the DMAC 209 is a word oriented (16 bit) device (A1 is its low order address bit) and since the local storage 210 is accessed by doubleword (32 bits), some means must be provided in the hardware to allow the DMAC 209 to read data into its internal link-list from contiguous local storage locations. This is accomplished by reading the same doubleword location in store 210 twice, using A2 as the low order address bit. Bit A1 is then used to select the high/low word from the local bus. The address bit shift to the local storage 210 is accomplished in the hardware via the DMAC function code bits. Any function code except "7" from the DMAC 209 will cause address bits A15 -A02 to be presented to the local storage 210. This scheme allows the local storage link list data for the DMAC 209 to be stored in contiguous locations in store 210.
  • the DMAC bit A1 is used is the low order address bit to the local storage 210.
  • the read data is supplied to storage 210 from the adapter bus Channel 0 read buffer 226. Data is written from storage 210 to the adapter bus Channel 1 write buffer 228. Since the DMAC is a 16 bit device, the low order address bit is intended to represent a word boundary. However, each DMAC operation accesses a doubleword. To allow for doubleword accesses with a word access addressing mechanism, an address shift is required.
  • the address bit shift to the local storage 210 is accomplished in the hardware via the DMAC function code bits.
  • a function code of 7 from the DMAC 209 will cause address bits A14 - A01 to be presented to the local storage 210.
  • the DMAC is loaded with 1/4 of the actual byte count (1/2 the actual word count).
  • word writes by controlling the UDS and LDS lines (not shown) from the DMAC 209, although all DMAC operations are normally doubleword accesses.
  • the UDS and LDS signals cause accessing of high (D31-D16) and low order portions (D15-DO) local store 210.
  • the S/88 processor PE 62 will write the DMAC registers in each of the four DMAC channels 0-3 in order to set up the internal controls for a DMAC operation.
  • PE 62 also has the capability of reading all of the DMAC registers.
  • the DMAC 209 returns a word (16 bit) DSACK on a bus 266 which has two lines DSACK 0, DSACK 1 permitting port sizes of 8, 16 or 32 bits. This allows the DMAC 209 to take as many cycles as necessary in order to perform the DMAC load properly.
  • the S/88 processor SIZO, SIZ1 (not shown) and AO lines are used to generate UDS (Upper Data Strobe) and LDS (Lower Data Strobe) inputs (not shown) to the DMAC 209. This is required in order to access byte wide registers in the DMAC 209 as described more fully in the above described DMAC publication.
  • the LDS line is generated from the logical OR of NOT SIZO or SIZ1 or AO of address bus 161 D.
  • the UDS line is generated from the logical NOT of A0.
  • the SIZO line is used to access the low order byte when a word wide register is being accessed (NOT SIZO).
  • the SIZ1 line is used to access the low-order byte when a word wide register is being accessed via a "three byte remaining" S/88 processor operation. This will only occur when the S/88 processor is performing a doubleword (32 bit) read/write operation to the DMAC on an odd-byte boundary.
  • Bit AO is used to select the high or low byte in a two-byte register.
  • Bits A0, A1 are used to select bytes in a four-byte DMAC register.
  • Bits A6, A7 of the PE62 address bus 161 D select one of the four DMAC channels.
  • the BCU 156 is capable of accepting a single command from the DMAC 209 which will transfer up to 4KB of data across each adapter BUS 250, 251. However, each bus can only handle 64 byte blocks for one data transfer operation. There are other after bus restrictions that must be obeyed by the hardware in order to meet the protocol requirements. The following is a detailed description of the BCU 156 hardware that accomplishes this.
  • the BCU 156 contains two fullword (11 bit) counters 220, 222 and two boundary (4 bit) counters 221, 224 that are used for adapter bus BSM read and BSM write operations.
  • the boundary counters 221, 224 are used to represent a starting address to bus adapter when a 64 byte boundary crossing is detected by the BCU 156 for any single command/data transfer operation, or when the byte count is greater than 64 bytes.
  • the boundary counter contents are presented to bus adapter 154 for all but the last block transfer.
  • the fullword counter contents are presented for the last block transfer only (last command/data transfer operation).
  • the S/88 processor 62 places byte count, key, and priority bits on the local bus 223 (Fig. 45F) for transfer to register 222 or 220.
  • the r bit (count bit 1) represents word (2 bytes) boundaries and the s bit (count bit 0) represents byte boundaries.
  • Fullword counter bits represent a 2KB-1 doubleword transfer capability. Since all transfers are done on a doubleword basis, bit 2 is the low order decrement bit.
  • the r and s bits are latched by the BCU and presented to bus adapter 154 on the final 64B transfer.
  • the S/88 processor will pre-calculate the double-word count and the r, s and i bits, based upon an examination of the factors described above, and the total byte transfer count. The r and s bits will not be presented to bus adapter 154 until the last command/data transfer operation.
  • the DMAC 209 captures bits 31-16, and BCU 156 captures bits 26-6.
  • BCU 156 stores bits 26 - 14 in register 220 or 222.
  • the bits 26 - 16 represent the doubleword count field.
  • Counter 220 or 222 is decremented on a doubleword boundary (Bit 2).
  • S/88 processor PE62 places a BSM Read/Write Select Up Command on the local address bus 247 and the BSM starting address on the local data bus 223.
  • the DMAC 209 is a 16 bit device which is connected to 32 bit bus. It is programmed to transfer words (2 bytes) during DMA operations in all channels, and each internal memory address register MAR increments by one word (2 bytes) per transfer. However, a double-word (4 byte) increment is required, since each transfer is actually 32 bits. To accomplish this, the S/88 processor PE62 always initializes the MAR to one-half the desired starting address (in store 210). The BCU 156 then compensates for this by doubling the address from the MAR before presenting it to the local bus 223, resulting in the correct address sequencing as seen at the store 210.
  • the BCU 156 performs the following:
  • the BCU 156 loads the BSM Read/Write command byte count to the command/status bus 249 or 252 from the boundary counter 221 or 224 and BSM address register 231 or 228 bits 1,0 (inverted). Then a Read/Write operation is performed.
  • the BCU 156 will decrement the boundary count registers 221 or 224 and the fullword count register 220 or 222 on a doubleword boundaries; in addition, it will increment BSM address register 231 or 228 on a doubleword boundary.
  • the BCU 156 When 64 bytes or less remain and there is no boundary crossing during a block transfer of data, the BCU 156 will load BSM Read/Write command byte count to adapter bus command/status bus 249 or 252 from bits 5-2 of counter 220 or 222 and the r, s bits. The BCU 156 then performs a Read/Write operation during which it decrements register 220 or 222 on a doubleword boundary, increments BSM address register 231 or 228 on a doubleword boundary, and stops when the register 220 or 222 bits 12-2 are all ones. A boundary crossing is detected by comparing bits 2-5 of count register 220 or 222 with its boundary register 221 or 224. If the count register 220, 222 value is greater than that of the boundary register 221, 224, then a boundary crossing has been detected.
  • the timing chart of Fig. 25 shows the handshaking sequences between the BCU 156 and the adapter 154 for Read Mailbox commands and storage Read commands including the transfer of two thirty-two bit words to a work queue buffer in local store 210.
  • a pair of signals Gate Left and Crate Right sequentially gate the left and right portions of the command and address in registers 214 and 219 (Fig. 19B) to adapter 154 to fetch the appropriate data from S/370 storage 162.
  • the Tag Up command is raised on line 262a followed by periodic Read Data signals.
  • Tag Down is raised on line 262b until the fetched data is stored in buffer 259.
  • the next of the periodic Clock Left and Clock Right signals are raised, the left and right portions of the first fetched word are gated into buffer 226 via bus 250.
  • Bus Request is raised on line 263a or b for DMAC channel 0 or 1.
  • DMAC arbitrates for control of the local bus via line 269. When this request is granted by logic 216, Bus Grant is raised on line 268.
  • DMAC 209 raises the Acknowledge signal on line 264a or 264b which causes the BCU to gate the data in buffer 226 to the local data bus 223 while DMAC 209 places the selected local store address on the local address bus 247.
  • the DMAC 209 then issues DTC on line 267 to cause logic 253 to raise the Store Select on line 210a; and the data on bus 223 is placed in the appropriate buffer in local store 210.
  • DMAC Request gate succeeding data words to buffer 226; and these words are transferred to the appropriate buffer in store 210 as DMAC 209 gains access to the local buses 247, 223 via arbitration logic 216 and issues Acknowledge and DTC signals.
  • Fig. 26 similarly shows the handshaking sequences for Queue Select Up and Storage Write Commands.
  • the Gate Left and Right signals transfer the command and address (previously stored in registers 225 and 227) to the adapter 154.
  • a Tag Up Command followed by periodic Data signals are raised on line 262a.
  • DMAC Request is raised on line 263c or d.
  • the DMAC 209 arbitrates for the local bus 247, 223 via line 269 and logic 216.
  • the DMAC 209 raises Acknowledge on line 264c or d followed by DTC on line 267 to transfer the first data word from store 210 to register 227.
  • the next periodic Gate Left and Right signals transfer the first data word from register 227 to the buffer 260 of adapter 154.
  • Succeeding DMAC Request signals on line 263c or d and DMAC Acknowledge and DTC signals transfer succeeding data words to register 227 as the DMAC 209 arbitrates for control of the local buses 247, 223.
  • Succeeding periodic Gate Left and Right signals transfer each data word from the register 227 to buffer 260.
  • Each processing ELEMENT such as PE85 of the preferred embodiment contains the basic facilities for the processing of S/370 instructions and contains the following facilities:
  • each processor element 85 of the preferred embodiment is a processor capable of executing the instructions of the System/370 architecture.
  • the processor fetches instructions and data from a real storage area 162 of the storage 16 over the processor bus 170.
  • This bi-directional bus 170 is the universal connection between PE85 and the other units of the S/370 chip set 150.
  • PE85 acts as mastr but has the lowest priority in the system.
  • the instructions are executed by hardware and by micro instructions which the processor executes when it is in micro mode.
  • PE85 has four major function groups:
  • the I-buffer 309 makes the S/370 instructions available to the decoder as fast as possible.
  • the first half word containing the op code is fed via operation register 310 to the decoder 312 to start the S/370 I-phase.
  • the second and third half words (if any) are fed to the ALU for address calculation.
  • the I-buffer 309 is a double word register which is loaded by operations such as IPL, LOAD PSW, or PSW swap via a forced operation (FOP) in register 313 prior to the start of a /370 instruction sequence.
  • the I-buffer 309 is refilled word-by-word as the instructions are fed to operation register 310 (and ALU 306, for address calculation), and it is refilled completely during each successful branch.
  • the operation decoder 312 selects which operation to perform.
  • the decoder is fed from the operation and the micro code operation registers 310. Mode bits decide which one (or none in case of a forced operation) gets control to decode.
  • the I-buffer 309 contents are fed into the operation register 310 and in parallel into the CSAR 308 to address an opcode table in the control store 171.
  • Each entry in this table serves two purposes: it indicates whether a microcode routine exists and it addresses the first instruction of that routine.
  • Microcode routines exist for the execution of the more complex instructions, such as variable field length instructions and all others that are not directly executed by hardware.
  • Special function codes in the micro instructions activate the supporting hardware so that it is possible to control the 32-bit data flow using mostly 16-bit micro instructions.
  • Effective processing is additionally enhanced by implementing the decoder in several groups (not shown), one specifically dedicated to the ALU, another to the bus group, and so forth.
  • Byte-selectable multiplexers (not shown) at the A/B register input and the ALU output further enhance the operations.
  • S/370 RR instructions which occupy each of the pipelining sages for only one cycle.
  • the forced operation registers (FOPs) 313 are used for internal control. They get input from traps and exceptional conditions, and force another mode into the decoder 312. Typical operations are I-buffer loading, transition to trap level, and the start of exception routines.
  • Each operation register 310 has a cycle counter 311 of its own.
  • the micro code cycle counters are shared by some forced operations (FOPs). The arithmetic operations and most of the other micro instructions require only one cycle. Most of the micro instructions which perform processor bus operations require two cycles.
  • the data local store 303 contains 48 full-word (4-byte) registers which are accessible via three ports, two being output ports, one being the input port. Any register can be addressed via register 314 for input, and the same register or two different registers can simultaneously be addressed for output. This three-fold addressing allows operand fetching to overlap with processing. Due to a comparator logic and data gating (not shown), a register just addressed for a write operation may also be used as input in the same cycle. This facilitates the pipelining actions.
  • the ALU 306 is preferably a full-word logic unit capable of executing AND, OR, XOR, and ADD operations in true and inverted form on two full word operands. Decimal addition is also supported. Parity prediction and generation as well as fast carry propagation is included.
  • the save register 320 supports divide operations. Status logic 321 generates and stores various conditions for branch decisions, sign evaluation, etc.
  • the control store address register (CSAR) 308 addresses micro instructions and tables in the control store 171.
  • the input to the CSAR 308 is either an updated address from the associated modifier 322 or a branch target address from a successful branch, or a forced address for a table look up.
  • a table look up is mandatory at the beginning of each S/370 instruction, and for some forced operations (FOPs).
  • the CSAR 308 gets the op code pattern as an address to access the op code table (Fig 29).
  • the output of this op code table defines the form of execution which may be direct decoding out of the operation register 310. If indirect execution is required, the op code table output is fed back into CSAR to address the appropriate micro routine.
  • the storage address register 302 is designed for 24-bit addresses.
  • An associated modifier 323 updates the address according to the size of the data block fetched. Instructions are fetched in advance in increments of one word (4 bytes) as the I-buffer 309 is being emptied.
  • the input to the storage address register 302 comes from the instruction operand address register 324. It is furthermore set in parallel with the instruction address register 324 for speed up reasons.
  • the CPU data flow allows the overlapped processing of up to three S/370 instructions at a time.
  • S/370 instructions are executed either in hardware or interpreted by microinstructions.
  • the basic cycle time of the preferred embodiment is 80 ns. Instruction processing is performed in one or more 80 ns steps.
  • a high speed multiply facility PE151 speeds up binary and floating point multiply operations.
  • Microinstructions from control store 171 are employed only for the execution of those S/370 instructions which are too complex and thus too expensive to he implemented entirely in hardware.
  • the microinstructions, if needed, are supplied at a rate of 60 ns per instruction.
  • the microinstruction set is optimized for the interpretation of S/370 instructions. Microinstructions have half word format and can address two operands.
  • Microcode not contained in the control store 171 is held in the IOA area 187 which is a reserved area in S/370 memory 162 (see Figs. 28, 29).
  • This microcode includes the less performance sensitive code for exceptions, infrequently executed S/370 instructions, etc.
  • These microroutines are fetched on a demand basis into a 64B buffer 186 in the RAM part of control store 171.
  • the PE85 encounters an address larger than implemented in the control store 171, it initiates a 64B block fetch operation to cache controller 153 and storage controller interface 155.
  • the units 153, 155 fetch the 64B block from the IOA 187 and send it to the PE85 which stores it into the buffer 186.
  • the microinstructions are fetched by PE85 from buffer 186 for execution. All microcode is loaded into memory at initial microcode load (IML) time.
  • the system provides an IML support to facilitate the microcode loading from the S/88 into the memory.
  • S/370 instructions and user data are fetched from an 8KB high speed cache storage 340 (Fig. 31). Data is read/written from/into the cache 340 on a full word basis. The time needed to read/write a full word from/into the cache is 120 nanoseconds. The cache 340 is automatically replenished with 64 byte blocks from the memory 162 when the need arises.
  • the PE85 communicates with the cache 340 via processor bus commands. The virtual addresses provided by the PE85 are used to look up the corresponding pretranslated page addresses in directory look aside table (DLAT) 341.
  • DLAT directory look aside table
  • the data local store 303 in PE85 includes 16 general registers, 4 floating point registers and 24 work registers. All registers can be addressed individually via three separately addressable ports. Thus the store 303 can feed two operands in parallel into the ALU 306 and simultaneously accept a full word from the ALU 306 or cache 340 within the same 80 ns cycle. Since there is no serialization as on conventional data local stores, arithmetic and logic operations can be executed in an overlapped manner with preparation for the next instructions.
  • the CPU maintains an 8-byte instruction buffer (I-Buffer) 309 for S/370 instructions. This buffer is initialized by a successful S/370 branch instruction.
  • the PE85 fetches a double-word of data from the S/370 instruction stream from cache 340 and loads it into the I-Buffer 309. When the first full-word is loaded in the I-Buffer 309, the PE85 starts instruction execution again. I-Buffer data is fetched from cache 340 simultaneously with the execution of S/370 instructions. Since the first cycle in each S/370 instruction execution is a non-cache cycle, the CPU utilizes this cycle for prefetching a full-word from cache 340 into the I-Buffer 309.
  • a second non-cache cycle is available with S/370 instructions which require indexing during the effective address calculation or which are executed by microroutines.
  • S/370 instruction fetching can be completely overlapped with the execution of S/370 instructions.
  • the S/370 chip set 150 communicates via an interrupt mechanism which requires the chip receiving an interrupt to acknowledge it by resetting the interrupt latch of the sending chip.
  • STR status register
  • N_ATTN_REQ N_ATTN_REQ control line
  • This causes an exception in the processor element 85 when the current S/370 instruction has been executed, thus forcing the processor element 85 to "look" at the status register.
  • An exception handler will then sense the STR contents, interrogate the 'interrupt type(s)' and dispatch the appropriate system microroutine(s).
  • the processor element 85 activates a bit in the STR, the system must react to it accordingly. Basically there are two types of interrupt requests:
  • the existence of two additional registers is assumed, the BR register 115 (Fig. 13) which is set by PE 85 and read by the system and the BS register 116 which is set by the system and read by PE 85.
  • the PE 85 sets data to be transmitted to the system into the register 115 and sets the PCR1 latch on.
  • the system reads the data from the register 115 and resets the PCR latch.
  • the processor 85 may sense the PCR latch to find out whether or not it has been reset.
  • the PE 85 may transfer further data to the system by repeating above sequence.
  • the system may transfer data to the PE 85 in a similar way as follows.
  • the system sets data to be transmitted to the PE 85 into the register 116 and sets the SCR latch on.
  • the PE 85 is interrupted, senses the STR, finds the SCR latch on, reads the data from the register 116, and resets the SCR latch.
  • the system may interrogate the SCR latch to find out whether or not it has been reset.
  • the system may transfer further data to the PE 85 by repeating above sequence.
  • Data can also be exchanged via the IOA storage area 187.
  • the PE 85 has one set of buffers assigned in the IOA area 187 into which it sets data to be fetched by the system.
  • the system has another set of buffers assigned in the IOA area 187 into which it sets data to be fetched by the PE 85.
  • the interrupt types IOASYS/ IOAPU may be used in SYSREQs to indicate to each other that data was set into IOA buffers.
  • PE 85 executes the following functions:
  • 1/0 interruption requests are generated by the system by setting the 1/0 bit in the STR.
  • the exception handler is invoked.
  • the PE 85 reads the STR to recognize the I/O interrupt request.
  • the PE 85 resets the STR bit and sets the interrupt request latch internal to the PE 85. This latch is masked with the I/O mask of the current PSW. If the mask is 1 and no higher priority interrupt requests are pending, the exception handler passes control to a system-provided 1/0 interrupt request handler which processes the 1/0 interrupt request.
  • the processor bus 170 is the common connection between all S/370 chip set components. Logically, all lines listed below belong to this bus:
  • the bus unit (PE 85, adapter 154 or cache controller 153) requesting control of the bus 171 sets the command on the bus. For CPU-storage and I/O-storage commands, the bus unit also sets the access key and dynamic address translation bit on the Key Status Bus. After completion of the command status is returned on the same bus to the requesting bus unit.
  • the adapter 154 issues CPU-storage commands and I/O-storage commands while PE 85 can only issue CPU-storage commands.
  • These command groups are as follows:
  • I/O-storage commands are executed in cache controller 153 without checking of the S/370 main storage address. This checking is performed in STCI 155.
  • CPU-storage commands are directed to controller 153 for execution and have a one byte command field and a three byte real or virtual address field.
  • the command field bits are as follows:
  • CPU-storage commands are:
  • I/O-storage commands are initiated by the adapter 154 and directed to the cache controller 153. They transfer data strings from 1-64 bytes in length in ascending address order.
  • the 32 bit command format includes a real byte address in the three low order bytes and the high order byte includes a highest order bit "0", next highest order bit defines a fetch or store operation and the remaining six bits define the length of the data transfer (1-64 bytes).
  • Data strings are transferred on word boundaries except for the fire and last transfer which may require position alignment on the bus.
  • MMU commands are used to control the cache controller 153 and its registers including DLAT, ACB, directory and the like.
  • Message commands are used to transfer messages between bus units connected to bus 151.
  • the cache controller includes the cache storage 340 and addressing and compare logic 347, 348, a fetch aligner 343, as well as the directory look-aside table (DLAT) 341 for fast address translation.
  • the controller 153 accepts virtual addresses and storage commands from the processor bus 170 and transfers fetch or store commands to the storage control interface 155 (Fig. 11) via multiplexer 349 and STC bus 157, when it cannot satisfy the request via cache storage 340.
  • DLAT 341 provides for fast translation of virtual page addresses into real page addresses. Its 2 x 32 entries hold 64 pretranslated page addresses.
  • the DLAT 341 is accessed using a 2-way set (associative addressing scheme.
  • the virtual page size is preferably 4KB.
  • the PE85 is interrupted and the virtual address translation is done by microprogram using segment and page tables (not shown) in S/370 main storage 162 in a well-known manner.
  • the DLAT 341 is then updated to reflect the new virtual and real page address of the information fetched from storage and placed into the cache. A copy of the storage key is fetched from the S/370 Key Storage and included into the DLAT entry.
  • the 8KB cache 340 with its associated cache directory 342 provides a high speed buffer to significantly improve the processor performance.
  • Data and directory arrays are partitioned into 4 compartments. Each compartment in the cache is organized 256 x 8B (bytes).
  • the byte offset in the virtual address is used to simultaneously address the DLAT 341, cache directory 342 and cache 340.
  • Key-controlled protection checking is done by compare circuit 345 using the storage key in the selected DLAT entry.
  • 4 x 8B of data are latched up at the output 340a of the cache 340. If the requested data is in cache 340, a late select signal is used to gate the appropriate bytes into the fetch aligner 343.
  • the cache controller 153 In case of a cache miss the cache controller 153 automatically sets up a BSM command to fetch the required 64B cache line in burst mode. If the cache line to be replaced by the new cache line was changed since it was loaded, a cache line cast-out operation to storage 162 is initiated before the new cache line is loaded. I/O data will never cause cache line cast-out and load operations. I/O data to be fetched from storage 162 will be looked for in both the main storage 162 and the cache storage 340 by accessing both facilities. If a cache-hit occurs, the memory operation is cancelled, and the cache storage supplies the data. If the I/O data is not in cache, it will be fetched directly from memory, but no cache line will be replaced. I/O data to be stored into storage will be stored into cache 340 if the addressed line is already in cache; otherwise, it will be stored directly into the storage 162.
  • the 4KB key storage 344 holds the storage keys for 16MB memory.
  • the key storage is an array organized 4K x 8. Each byte holds one storage key. Each DLAT entry holds a copy of the storage key associated with its 4KB-block address. This reduces significantly the number of accesses to the key storage while repetitively accessing a page. Changes in storage key assignments affect both the key storage and any copies in cache storage.
  • Commands, data and addresses received by the cache controller 153 from the processor bus 170 via receiver circuit 355 are stored in the command, data and address registers 350, 351 and 352.
  • Address register 347 stores the range of valid address- for the related S/370 processing element PE85.
  • the compare logic 348 verifies the validity of the received address.
  • the S/370 address compare function provided by address register 347 and its related compare logic 348 handles addresses from both the PE85 and the I/O bus adapter 154.
  • the Address Compare Boundary (ACB) register 353 compare function ensures that S/370 main storage references intended for the customer area do not address the IOA area.
  • the ACB register 353 stores the dividing line (boundary) between the reserved IOA area and the non-reserved area in S/370 storage 162. Each access to S/370 storage results in compare logic 354 comparing the received address with the ACB value.
  • the storage control interface (STCI) 155 connects the S/370 chip set 150 to tile S/88 duplexed fault-tolerant storage 16, 18 via bus logic 178 and the system bus 30 (Fig. 1). It supports all S/370 processor and 1/0 store/fetch commands which define data transfers from 1-64 bytes per command. All ECC, refresh. memory initialization and configuration, retries, etc. are handled by S/88 processor 62 and storage 16, 18. A detailed dataflow of the STCI 155 is shown in Figs. 32A, B.
  • arbitration such as logic 408 (Fig. 32B) in each STCI.
  • STCI 155 must arbitrate against its associated S/88 processor 62 (and that processor's paired and partnered processors in CPUs 21, 23 of Fig. 8) which may be requesting control of the bus for S/370 1/0 functions or conventional S/88 functions.
  • any unit of the processor module 9 which is capable of being a bus master and which is ready to initiate a bus cycle, arbitrates for use of the bus structure.
  • the unit does this by asserting a Bus Cycle Request signal and by simultaneously checking, by way of an arbitration network, for units of higher priority which also are asserting a Bus Cycle Request.
  • the unit, or pair of partnered units, which succeeds in gaining access to the bus structure during the arbitration phase is termed the bus master and starts a transfer cycle during the next clock phase.
  • Each memory unit 16, 18 is never a master and does not arbitrate.
  • the unit which is determined to be the bus master for the cycle defines the type of cycle by producing a set of cycle definition or function signals.
  • the bus master also asserts the address signals and places on the address parity line even parity for the address and function signals. All units of the processor module, regardless of their internal operating state, always receive the signals on the bus conductors which carry the function and address signals, although peripheral control units can operate without receiving parity signals.
  • the cycle being defined is aborted if a Bus Wait signal is asserted at this time.
  • any addressed unit of the system which is busy may assert the Bus Busy signal to abort the cycle.
  • a memory unit for example, can assert a Bus Busy signal if addressed when busy or during a refresh cycle.
  • a Bus Error signal asserted during the response phase will abort the cycle, as the error may have been with the address given during the definition phase of the cycle.
  • Data is transferred on both the A bus and the B bus during the data transfer phase for both read and write cycles. This enables the system to pipeline a mixture of read cycles and write cycles on the bus structure without recourse to re-arbitration for use of the data lines and without having to tag data as to the source unit or the destination unit.
  • Full-word transfers are accompanied by assertion of both UDS and LDS (upper and lower data strobe) signals.
  • Half-word or byte transfers are defined as transfers accompanied by assertion of only one of these strobe signals.
  • Write transfers can be aborted early in the cycle by the bus master by merely asserting neither strobe signal.
  • Slave units, which are being read, must assert the strobe signals with the data.
  • the strobe signals are included in computing bus data parity.
  • the normal backplane mode of operation of the illustrated system is when all units are in the Obey Both mode, in which both the A bus and the B bus appear to be free of error.
  • all units In response to an error on the A bus, for example all units synchronously switch to the Obey B mode.
  • the module 9 returns to the Obey Both mode of operation by means of supervisor software running in a S/88 central processing unit.
  • both the A bus and the B bus are driven by the system units and all units still perform full error checking.
  • the only difference from operation in the Obey Both mode is that the units merely log further errors on the one bus that is not being obeyed, without requiring data to be repeated and without aborting any cycles.
  • a Bus Error signal however on the obeyed bus is handled as above and causes all units to switch to obey the other bus.
  • Fig. 33 illustrates the foregoing operation with four pipelined multiple-phase transfer cycles on the bus structure 30 for the module 9.
  • Waveforms 56a and 56b show the S/88 master clock and master synchronization signals which the clock 38 applies to the X bus 46, for twenty-one successive timing phases numbered (1) to (21) as labeled at the top of the drawing.
  • the arbitration signals on the bus structure, represented with waveforms 58a change at the start of each timing phase to initiate, in each of the twenty-one illustrated phases, arbitration for a new cycle as noted with the cycle-numbering legend #1, #2, #3...#21.
  • Fig. 33 represents the cycle definition signals with waveform 58b.
  • the cycle definition signals for each cycle occur one clock phase later than the arbitration signals for that cycle, as noted with the cycle numbers on the waveform 58b.
  • the drawing further represents the Busy, Wait, Data, A Bus Error, and B Bus Error signals.
  • the bottom row of the drawing indicates the backplane mode in which the system is operating and shows transitions between different modes.
  • the module 9 produces the cycle arbitration signals for cycle #1.
  • the system is operating in the Obey Both mode as designated.
  • the Bus Master unit determined during the cycle arbitration of phase (1) defines the cycle to be performed during timing phase (2), as designated with the legend #1 on the cycle definition signal waveform 58b. Also in timing phase (2), the arbitration for a second cycle, cycle #2, is performed.
  • timing phase (3) there is no response signal on the bus structure for cycle #1, which indicates that this cycle is ready to proceed with a data transfer as occurs during timing phase (4) and as designated with the #1 legend on the data wave form 58e. Also during timing phase (3), the cycle definition for cycle #2 is performed and arbitration for a further cycle #3 is performed.
  • timing phase (4) the data for cycle #1 is transferred, and the definition for cycle #3 is performed. Also, a Bus A Error is asserted during this timing phase as designated with waveform 58f. The error signal aborts cycle #2 and switches all units in the module to the Obey B mode.
  • the Bus A Error signal of timing phase (4) indicates that in the prior timing phase (3) at least one unit of the system detected an error regarding signals from the A bus 42. The error occurred when no data was on the bus structure, as indicated by the absence of data in waveform 58e during timing phase (3), and there hence is no need to repeat a data transfer.
  • timing phase 5 With the system operating in the Obey B mode, a fifth cycle is arbitrated, the function for cycle #4 is defined and no response signal is present on the bus structure for cycle #3. Accordingly that cycle proceeds to transfer data during time phase (6). Also in time phase (6), a Bus Wait is asserted, as appears in waveform 58d; this is in connection with cycle #4. The effect is to extend that cycle for another timing phase and to abort cycle #5.
  • a new cycle #7 is arbitrated in timing phase (7) and the definition operation proceeds for cycle #6.
  • the data for cycle #4 is applied to the bus structure for transfer.
  • a Busy signal is asserted, This signal is part of the response for cycle #6 and aborts that cycle.
  • the Bus Wait signal asserted in time phase (10) and continuing to time phase (11) extends cycle #8 for two further time phases, so that the data for that cycle is transferred during time phase (13), as designated.
  • the Bus Wait signal asserted during these phases also aborts cycles #9 and #10, as shown. Any Busy signal asserted during phase (10), (11) or (12) in view of the extension of cycle #8 by the Wait signal, would abort cycle #8. Note that the data transfer for cycle #7 occurs in time phase (10) independent of the signals on the Wait and the Busy conductors during this time phase.
  • Bus Wait signal is driven only by slave units which have been addressed by a bus master unit and are not ready to effect a data transfer. Since the STCI 155 is never a slave unit and only addresses memory, not I/O devices, this line is not utilized by the STCI 155.
  • the system bus logic 178 (Fig. 19C) provides the link from the STCI 155 to the S/88 memory boards 16, 18 and includes arbitration logic 408 (Fig. 32B). The same basic transfer cycles defined above for the bus 30 are used by logic 178:
  • S/88 processor 62 arbitrating for the bus 30 and STCI 155 arbitrating for the bus 30 may now be described.
  • a S/88 processor 62 will be operated in only one of the five phases at any moment in time.
  • the STCI ran operate in up to all five phases at the same time. For example, during a 64 byte read operation, STCI 155 can be operated in all five phases at the same time if there are no errors and STCI is granted arbitration control of the bus 30 in each of five succeeding cycles. This improves system performance, especially in a uniprocessor version of a module 9.
  • AFE Almost Full/Almost Empty
  • HF Half Full
  • the memory accepts 9-bit parallel words at its inputs under the control of the Shift-In (SI) input when the Input-Ready (IR) control signal is high.
  • the data is output in the same order as it was stored under the control of the Shift-Out (SO) input when the Output-Ready (OR) control signal is high. If the FIFO is full (IR low) pulses at the SI input are ignored; if the FIFO is empty (OR low) pulses at the SO input are ignored.
  • Parallel expansion for wider words is implemented by logically ANDing the IR and OR outputs (respectively) of the individual FIFOs together.
  • the AND operation insures that all of the FIFOs are either ready to accept more data (IR high) or are ready to output data (OR high) and thus compensate for variations in propagation delay times between devices.
  • the FIFO 400 includes a write pointer, a read pointer, and the control logic necessary to generate known handshaking (SI/IR, SO/OR) signals as well as the Almost Full/Almost Empty (AFE) and the Half Full (HF) flags.
  • SI/IR, SO/OR handshaking
  • AFE Almost Full/Almost Empty
  • HF Half Full
  • the data is not physically propagated through the memory.
  • the read and write pointers are incremented instead of moving the data.
  • the time required to increment the write pointer and propagate a signal from the SI input to the OR output of an empty FIFO (fallthrough time) or the time required to increment the read pointer and propagate a signal from the SO input to the IR output of a full FIFO (bubblethrough time) determine the rate at which data can be passed through FIFO 400.
  • the FIFO Upon power up the FIFO is reset with a Master Reset signal. This causes the device to enter the empty condition, which is signified by the OR signal being low at the same time that the IR signal is high. In this condition, the data outputs (D00.D08) will be low. The AFE flag will be high and the HF flag will be low.
  • IR Input Ready
  • SI Shift-in
  • AFE Almost Full/Almost Empty
  • HF Half Full
  • SBI logic - System/88 Bus Interface (SBI) logic 178 which allows S/370 processor 85 to initiate read/writes to S/88 storage 16. It includes logic 408 to arbitrate every cycle for access to the bus 30 to initiate 16, 32, or 64-bit transfers.
  • the logic 178 interface lines and the arbitration logic 408 are preferably of the type described in the Reid patent to the extent that they are not modified as described herein.
  • STCI 155 has a substantially identical paired STCI 155a (not shown) which is a part of the storage management unit 83 of Fig. 8.
  • the comparator logic 402 a-g forms the compare logic 15 of Fig. 8 and broken logic 403 forms a part of the common control logic 75 of Fig. 8.
  • S/370 compare checking is performed only at the paired STCls 155, 155a to protect against dispersion of erroneous data via bus structure 30.
  • S/370 machine check and parity errors are supplied to logic 403 via bus 460.
  • Some errors on BCU buses 247, 223 are picked up by S/88 compare circuits 12f (Fig. 8).
  • Synchronous operation - S/370 clocks 152 are derived from the S/88 clock 38 (Fig. 7) 16 Mhz input, via bus 30 and synchronizing logic 158 (Fig. 19C), to allow synchronization between the clocks within one S/370 oscillator input clock period from the start of the S/88 clock 38.
  • This allows consecutive reads (eg a 64-byte read command) to be pipelined from memory 162 to the S/370 chip set 150 with no wait states in between (assuming consecutive cycles granted to STCI 155 on the system bus 30).
  • STC Bus interface All standard S/370 fetch/store commands are executed along with command cancelling. Parity errors and/or ECC errors will not be reported to the S/370 operating system but rather handled as retries (ECC or bus parity errors) or going broken (internal board parity errors). 64-byte line boundary crossings will result in address wraparound.
  • the STCI 155 interfaces to the S/370 processor 85 via the cache controller unit 153 which handles S/370 dynamic (virtual) address translation, utilizing an 8KB instruction/data cache 340 as well as a 64-entry DLAT 341 (directory lookaside table).
  • S/370 dynamic (virtual) address translation utilizing an 8KB instruction/data cache 340 as well as a 64-entry DLAT 341 (directory lookaside table).
  • all real/virtual I/O or processor transfers result in a 'real' address issued on the STC Bus 157 by unit 153.
  • unit 153 simply acts as a transition stage from the processor bus 170 to the STC Bus 157, except for cache hits which may result in a command being cancelled after having been issued on the STC Bus 157.
  • STC data/address/command bus 406 has 32 bidirectional data bus lines plus odd parity per byte. This bus is used to convey command and address in one cycle, and up to 32 bits of data on each subsequent cycle of the storage operation.
  • STC Valid line is driven by unit 153 to STCI 155 to signal that a command/address is valid on the STC Bus in the same cycle.
  • STC Cancel line is driven by unit 153 to STCI 155 to cancel a previously issued command. It may appear up to 2 cycles after STC Valid is issued. It is ORed with the PE 85 command cancel input.
  • STC Busy line 440 is driven by STCI 155 to unit 153, one cycle after an 'STC Valid' is issued, to signify that the unit is busy and can't accept a new command. It is released 1 cycle before the unit 155 is able to receive a new command.
  • STC Data Invalid on line 433 may be issued by the STCI 155 to unit 153 in the same cycle as data is returned on a fetch to invalidate the data transfer.
  • Unit 153 ignores the data cycle if the line is activated. This line will be sent coincident with data when a Fast ECC error has occurred on bus 30, data has miscompared between the logic of paired STCI units 155, 155a or incorrect parity was detected during a bus 30 read cycle.
  • STC Data Transfer line 441 is driven to unit 153 by the STCI 155 to signal a data transfer on the STC Bus 157 in the subsequent cycle. For stores, it dictates that unit 153 supply the next 32-bit word on the following cycle. For fetches, it alerts unit 153 that the next cycle will contain valid data, unless overridden by STC Data Invalid on next cycle.
  • the STCI 155 design is fully pipelined to allow all the above states to be active at the same moment within one S/370 CPU. In this fashion, assuming continuous bus grants and no bus errors, the STCI 155 can maintain pipelined data on fetches with no wait states utilizing 64-bit reads (per 125 ns system bus 30 cycle) onto the 32 bit, 62.5 ns STC Bus 157.
  • the System/88 interface 410 is used in STCI 155 to support access to the MEM Size/MEM Base registers 405 and 404 within the BCU local virtual address space. Also 'Broken' 403 and 'Bus Interrupt Req' (IRQ) errors are merged with those on the S/88 processor board 102 to drive a low priority maintenance interrupt on the bus 30 as a single CPU.
  • IRQ Interrupt Req'
  • Bus IRQ errors differ from broken in that these errors, usually due to unprotected signals from bus 30 which are detected different by same or partner board, do not disconnect a board from bus 30 as does broken. These errors are only active when the board is in Obey Both mode.
  • 'Obey A', 'Obey B', and 'Duplexed' signals on lines 411, 412, 413 are driven up from S/88 processor board logic 415 rather than reimplementing within the S/370 processors.
  • Obey A/Obey B signals are used to control the input multiplexors 71, 73 (Fig. 8) for the check and drive side data input multiplexors respectively, as well for gating in Bus error conditions.
  • the duplexed signal on line 413 is used for signalling when boards are partnered (i.e. used in bus arbitration logic 408 for ensuring both partners arbitrate together when in consecutive slots).
  • Obey A and B signals are inverted to provide both + Obey A, -Obey A, + Obey B and -Obey B.
  • the + Obey A and -Obey A signals are applied to registers 428 and 429 respectively.
  • Registers 428 and 429 are coupled respectively to the A and B buses of bus structure 30 respectively.
  • S/88 clock signals (not shown) clock data from the A and B buses to registers 428 and 429 respectively for all three clock modes A, B, and Both.
  • Data in register 428 is gated out on buses 435, 436 when the bus is operating in an Obey A or Obey Both modes and register 429 is gated out on buses 435, 436 only during the Obey B mode.
  • register 428a of STCI 155a are similarly gated out during Obey B or Obey Both modes.
  • the contents of register 429a are gated out during Obey A mode.
  • Dot ORing of the outputs of registers 428, 429 and 428a, 429a performs the respective data input multiplexer functions 71, 73 (Fig. 3).
  • the MEM Size/MEM Base values in registers 405, 404 are memory-mapped in the S/88 processor 62 virtual address space, by way of the BCU local address space. They must be set during the S/88 boot process once the given S/370 CPU space 162 is defined. They can be altered by the S/88 as long as no STCI store/fetch operations are in process.
  • the registers 404, 405 are accessed by the address decode logic 216 of Fig. 19A via a local address (007E01 FC) and include the following data: PA bits 20-23 and PA bits 20-27 which equal respectively the S/370 storage 162 size (MEM size) and storage base address (MEM Base) where:
  • logic 216 When logic 216 decodes the address 007E01 FC, the size and base address bits are set in registers 405, 404 by processor 62 via its bus 161 D. During this operation, logic 216 uncouples the processor 62 from its associated hardware, whereby the loading of registers 404, 405 is transparent to the S/88 operating system. In addition, the S/370 operating system is unaware of their existence or their use in accessing the S/370 storage 162.
  • Figs. 32A, B and 30 also illustrate signal 1/0 lines used by the storage control interface 155. This includes in addition to the STC Bus 157 all lines required to interface to the S/88 system bus 30, the S/88 processor 62 and the logic 415 on S/88 CPU board 102. For ease of description, the transceivers 13 of Fig. 8 are not shown in Figs. 32A, B.
  • the STCI 155 On a store command from cache controller unit 153, the STCI 155 will clock the command in on address/data bus 406 (which is part of STC bus 157) bits 0-7 and store it in the command buffer 416 along with the STC Valid bit and in buffer 417. STC Busy will be raised on line 440 during the next cycle by logic 401 to indicate that the unit 155 is busy. Meanwhile the 24-bit real address on bus 406 is also clocked into the A/D register 417.
  • STC Data Transfer will be raised by logic 401 and will remain active every cycle until all STC Bus data transfers for this command are complete. On stores, STC Data Transfer is not issued (and thus the command is not shifted into FIFO) until it is assured no cancel has been issued (up to 2 cycles after STC Valid). However, during this time logic 401 shifts the 24-bit address from register 417 to register 442 and the first four bytes of data are transferred from unit 153 to register 417. In addition, the FIFO HHF and AFE flags 409 are compared to the byte transfer length decoded from command buffer 416.
  • the FIFO flags indicate 1 of 4 ranges of buffer depth in use. If the byte transfer length plus the 4 bytes of command word data exceed the FIFO 64 word capacity when added to the worst case buffer depth, as indicated by the FIFO flags, then all STC Data Transfer activations are held up until this overflow condition disappears. This will occur as soon as enough words are shifted out of the FIFO to cause a change in the flag status.
  • command decodes from block 401 concatenated with the 24-bit address from register 442, via multiplexer 447, are stored in FIFO 400.
  • Subsequent 32-bit data blocks from A/D register 417 are stored in FIFO 400 in consecutive cycles, via register 442, once the initial store command is shifted into the FIFO.
  • Gate 423 is used to multiplex the lower 16 bits onto the upper 16 bits, for 16 bit transfers onto bus 30.
  • the S bit is used to distinguish stores from fetches and the C/A bit is used to differentiate between command words and data words in FIFO 400 as seen in Fig. 35. Parity is maintained through the FIFO.
  • the FIFO inputs and outputs are clocked differently. Data is shifted into the FIFO 400 with S/370 clocks, while being shifted out with S/88 clocks. The timings are set to allow for worst case fallthrough time of FIFOs (60 ns) when FIFO 400 is empty.
  • Individual sequencers in block 401 on the input/output sides of the FIFO 400 track transfers in/out of the FIFO.
  • the output sequencer actually tracks the number of bus 30 data transfers pending for the current fetch or store command.
  • the arbitration logic 408 is set to begin arbitration.
  • Cycle control logic in 408 will track all active STCI 155 bus 30 phases for both fetch and store operations. Together with bus 30 status lines (i.e., Bus Busy, Bus Error) this logic is used within STCI 155 to handle normal bus 30 phase operations as well as for handling error conditions resulting in cancelled cycle definition or data phases.
  • bus 30 status lines i.e., Bus Busy, Bus Error
  • the physical address is formed by first comparing in logic 422 the upper four bits of the S/370 24-bit real address from the FIFO 400 with the S/370 storage size value in register 405. If the S/370 address bits do not exceed the size region allotted for the S/370 processor 85, the upper four bits are then added by logic 423 to the S/370 storage base value in register 404, and concatenated to lower bits 19-1 in buffer 420 to form a physical 27-bit word address which is used as the starting S/88 address into the S/370 area 162. Otherwise a soft program check is reported. Any 64-byte address boundary crossings will result in wraparound to the starting address.
  • the address U/D counter 421 is used to hold bits 5-2 of the outgoing physical address. It is clocked in synchronization with the output sequencer, and while normally incremented, may be decremented when responding to Bus Busy or Bus Error conditions of a cycle response phase. Once the output sequencer is loaded, associated logic initiates store cycles based on bus arbitration grants via logic 408 while responding to Bus Error and Bus Busy conditions. An appropriate S/88 function code is produced by logic 401 corresponding to the S/88 store command; and the function code is placed in register 443 for application to the A, B buses of bus structure 30 when an arbitration cycle request is granted.
  • the output sequencer is normally decremented on each grant, by one for 32-bit and by two for 64-bit transfers to bus 30, until it reaches zero, indicating no further bytes are to be transferred by the present command.
  • the output sequencer will be incremented by one for cancelled 32-bit transfers and by two for 64-bit transfers (fetch only).
  • the address U/D counter 421 is decremented by one for cancelled 32-bit transfers and by two for 64-bit transfers (fetch only).
  • the data out register 425 is used to buf far outgoing data.
  • the data out hold register 426 is required in the event data must be redriven because of a subsequent Bus Error (A or B bus).
  • subsequent data (to a higher address) may be accepted and stored in storage 16, 18 earlier than the previous cycle data which is associated with the Bus Error because that data transfer must be repeated 2 cycles after its initial transfer. (Unlike stores, fetched data cannot be received out of sequence.)
  • the Bus Arbitration logic 408 arbitrates continuously for cycles until all transfers have been initiated and accepted on the bus 30.
  • the arbitration and data transfer to system bus 30 and store 16, 18 are similar to those previously described in section (b).
  • the FIFO desigm allows the storage of up to 64 words (almost 4 groups of 64-byte store transfers) before going busy. For stores, as long as the FIFO is not full and can accept the command and data words associated with the store, the FIFO is loaded continuously until done. Consequently, STC Busy is dropped after each store command is executed, releasing .unit 153 and allowing the S/370 processor 85 to continue execution. Assuming a high cache hit ratio in unit 153, performance is improved significantly by buffering the equivalent of almost four 64-byte stores in the FIFO or thirty-two 1-4 byte stores.
  • STCI 155 is the "drive” side of the STCI pair 155, 155a and that STCI 155a is the error "check” side. Therefore, only STCI 155 drives signals (control, address, data) onto the bus structure 30 as shown in Fig. 32B. Where signals are intended for both buses A and B, the STCI 155 drive lines are shown coupled to both buses (through the transceivers 13 not shown in Fig. 32B). In STCI 155a, the corresponding lines are not coupled to the bus structure 30; merely to the compare logic 402a-g.
  • Compare logic 402g compares address bits 27-6 from buffer 420, address bits 5-2 from address U/D counter 421, modified address bit 1 and the parity bit from parity generator logic 445, and the function code from register 443 with corresponding bits from STCI 155a. In the event of a miscompare, logic 402g applies error signals to the broken logic 403 and to Bus Error A and B lines.
  • Logic 402e compares data out bits from data out register 425 with corresponding bits from STCI 155a and applies miscompare signals to logic 403 and to Bus Error A and B lines.
  • Logic 402d compares bits from FIFO logic 401 with corresponding bits from STCI 155a. AND gate 446 provides an error signal to logic 403 if the STC Valid signal is raised while the STC Busy signal is active on line 440.
  • a fetch command follows the same path as store commands through registers 416, 417, 442 and the FIFO 400 as described above.
  • the STC Data Transfer signal is not raised on the STC Bus logic 408 until data is known to be received in register 428 or 429 from storage 162 via the bus 30.
  • a fetch command and an STC Valid signal are received and stored in register 416.
  • the command and its initial storage address are stored in register 417.
  • the STC Bus logic in 401 issues an STC Busy signal during the next STC Bus cycle to prevent the cache controller 153 from sending another command until STC Busy is removed.
  • the STC Busy signal is maintained by logic 401 until the fetch command is fully executed because the cache controller 153 is waiting for the fetch data to be received. (During store cycles STC Busy was removed as soon as all store data was transferred from the controller 153 to the FIFO 400.) During a fetch command cycle, STC Busy must he maintained until any and all store commands in the FIFO 400 are executed, then the fetch command is executed. Only then can STC Busy be removed to permit transfer of the next command to the STCI 155.
  • the STCI 155 will enter the data phase assuming no bus busy or bus error was reported during the cycle response phase.
  • the first 32 bits along with bits DP, UDS, and LDS are received on the A,B buses of structure 30 from the appropriate location in area 162 of storage 16 and partner, and latched into registers 428, 429 respectively, with the S/88 clock beginning the second half of the bus 30 cycle.
  • data will be gated from register 428 onto buffer 430 in the next S/88 clock cycle (start of next bus 30 cycle).
  • the second 32 bits are latched into registers 428 & 429 concurrently with the transfer of previous data to buffer 430.
  • a parity generator 431 adds odd parity to the data word stored in 430. These data and parity bits, along with the UDS, LDS, and DP bits received, are applied to logic 402c via buses 435 and 436. Logic 402c compares these bits with the corresponding bits produced in the paired STCI 155a. Buffer 430 will now gate the first data word, plus parity, onto buffer 432 to be driven during the next STC bus cycle for transfer to cache controller 153 via bus 406 of STC bus 157. Buffer 432 is clocked with S/370 clocks which are synchronized with S/88 clocks such that the beginning of the STC bus cycle occurs after activation of the S/88 clock.
  • STC Data Invalid is issued on line 433 by logic 402c concurrently with the data on the STC address/data bus 406. Furthermore, if subsequent data arrives in the cycle after the cycle in which data is invalidated, a Bus error condition will be forced by the STCI SBI logic on both A and B buses following that data cycle. This ensures that data will be redriven 2 cycles later (i.e. one cycle after Bus error is reported), thus maintaining data integrity and functionality on the STC Bus by transferring fetched data in sequence.
  • Driving bus errors on both A and B buses is equivalent to memory 16 reporting an ECC error condition versus a 'true' bus error, thus not causing a change in bus OBEY logic along all controllers on the system bus 30.
  • the same logic 402c used to compare incoming data and check parity via buses 435, 436 is also used on store operations to verify the results of the data output comparison in 402e by performing a 'loopback' data comparison from the system bus 30 via register 428 or 429.
  • This helps identify transceiver 13 problems on the board 101 faster and will set the board broken logic 403 on stores if there is a miscompare and a bus error is not reported in the next bus cycle.
  • all comparator outputs 402a-g which produce a fault condition on valid miscompares for fetch and store operations, will generate a broken condition in logic 403. The initial setting of broken will generate bus error signals on both A and B buses, thus ensuring that a data transfer in the previous cycle is repeated, while any cycle definition phase in the previous cycle is aborted.
  • 64-bit writes are not available in the preferred embodiment of unit 155 due to the emphasis placed on minimizing hardware.
  • a 64x36 FIFO is sufficient to support 32-bit store transfers from S/370.
  • One performance limitation resulting from using only 32-bit writes is that since each S/88 memory board 'leaf' in interleaved storage 16 is 72 bits long (64 bits plus 8 ECC bits), each leaf, once accessed on writes, will stay busy for three (3) additional (125 ns) cycles. This means that the same leaf can be accessed only once every 5 cycles (625 ns) on consecutive writes.
  • Each 27-bit address and each 4-bit function code are sent together with an accompanying parity bit during bus 30 cycle definition phases.
  • the 32-bit data also carries a parity bit associated with it during bus 30 data phases.
  • a basic 125 ns cycle on bus 30 allows for normal 16 and 32 bit transfers, as well as 64-bit read transfers within the 125 ns window.
  • additional hardware can be used to support consecutive 64-bit write transfers in STCI 155.
  • Fig. 37 illustrates diagrammatically an overview of the S/88 hardware and application code which is utilized to support S/370 1/0 functions.
  • the hardware devices are 601, 602, 615-619, 621 and 623-625.
  • the software (or firmware) routines are 603-614, 620, 622 and 626.
  • Block 606 is the main control for the S/88 application code which consists of Block 606 through Block 614.
  • This set of blocks known as EXEC370, performs all the S/88 application code functions pertaining to the emulation and support of S/370 external devices, services, configuration, operator's console, etc.
  • Block 603 is the microcode running in the S/370 microprocessor. It supports the S/370 CPU functions. A protocol between Block 603 and Block 606 enables them to communicate requests and responses with each other regarding the initiation of S/370 1/0 operations, their completion, and S/370 1/0 device and channel status information. It also enables Block 606 to request Block 603 to perform specific S/370 CPU functions.
  • Block 605 is S/370 storage, and it is directly accessible to both Block 603 and Block 606.
  • Block 606 provides the proper S/370 configuration via the data contained in Block 602 which is a S/88 data file.
  • Block 604 is a separate running task which provides the S/370 operator's panel through a S/88 terminal device. This task may be started or stopped at any time without disrupting the logical functioning of the S/370 process.
  • Block 607 is a part of EXEC370 and provides interface emulation function between the S/370 process and Block 604.
  • Block 601 is a set of S/88 data "patch files" containing S/370 object code which has been written especially for the purpose of debugging the S/370 including its BCU 156. There is a debug panel provided by Block 604 which allows for the selection and loading into Block 605 of one of these "patch files.”
  • Block 608-1 consists of the code responsible for emulating the S/370 channel. It performs the fetching of S/370 CCW's, the movement of data to and from Block 605, the reporting of S/370 I/O interrupt information to Block 603, and the selection of the proper Control Unit code emulator. There may be more than one S/370 channel (e.g., 608-2), however the same code is used.
  • Block 609-1 is the S/370 Control Unit emulator code.
  • System 370 has many different types of control units, i.e., DASD controllers, tape controllers, communication controllers, etc.
  • the S/370 controller function is partitioned between Block 609-1 and the particular device emulator, Block 610 through Block 614.
  • the major purpose of Block 609-1 is address separation functions, however other Control Unit specific functions may reside in Block 609-1.
  • There therefore is more than one block of this type e.g., Block 609-2), i.e., DASD controller emulator, communications controller emulator, etc.; but there is not a one to one correspondence with those S/370 Control Units supported.
  • Block 610 represents the code necessary for emulating a S/370 console.
  • Block 611 represents the code necessary for emulating a S/370 terminal.
  • Block 612 represents the code necessary for emulating a S/370 reader. This is a virtual input device patterned after the standard VM reader. It provides for the input of sequential files which have been generated from another source, typically tape or diskette.
  • Block 613 represents the code necessary for emulating a S/370 printer.
  • An actual S/88 printer may be driven or the S/370 data may written to a S/88 file for spool printing later.
  • Block 614 represents the code necessary for emulating a S/370 disk.
  • the two formats: Count, Key and Data; and Fixed Block are supported by two different sets of code.
  • Block 615 represents a S/88 terminal, typically the S/88 console output device.
  • the System/88 console displays both S/88 operator messages and S/370 operator messages in addition to logging the messages to a log on disk which will appear to the S/370 is a 3278 or 3279 terminal.
  • Block 616 represents a S/88 terminal.
  • Block 617 represents a S/88 sequential data file on a S/88 disk.
  • Block 618 represents a S/88 printer or a sequential data file on a S/88 disk.
  • Block 619 represents a S/88 data file on a S/88 disk.
  • Block 620 is the code which will read a System/370 tape mounted on a S/88 tape device, and format it into Block 617 as it appears on the original S/370 tape.
  • Block 621 represents a S/88 tape drive with a S/370 written tape mounted.
  • Block 622 is the code which will read a file entered into S/88 from a Personal Computer, and format it into Block 617 as it originally appeared when it was generated on a S/370 System.
  • Block 623 is a Personal Computer configured to send to and receive data from both a S/88 and a System/370.
  • Block 624 is a S/370 System.
  • Block 625 represents a S/88 spooled printer.
  • Block 626 is the code which formats a S/88 file into an emulated System/370 DASD device. This is a S/88 separately run task which will format the file to any of the supported S/370 DASDs desired.
  • S/370 Architecture provides several types of 1/0 instructions, a program-testable condition-code (CC) scheme, and a program interrupt mechanism.
  • CC program-testable condition-code
  • an I/O instruction is directed toward an 'I/O Channel', which directs and controls the work of the I/O operation in parallel with other CPU processing, and reports status to the CPU when the I/O instruction is executed (via condition-code), and/or when the I/O operation is completed (via program interrupt).
  • the broad view of the Fault Tolerant System/370 improvement is then a S/370 CPU (chipset with customized firmware) and a 'pseudo-I/O-Channel' consisting of time-slices of a S/88 CPU and Operating System (OS), with the addition of special firmware and application-level software (EXEC370) providing both S/370 I/O device emulation and overall control of the system complex.
  • S/88 portion of this complex provides fault-tolerant CPU, OS, I/O devices, power/packaging, busses, and memory; the S/370 CPU is made fault-tolerant through hardware redundancy and added comparison logic.
  • the required custom firmware (i.e., microcode) falls into two groups:
  • a broad view of the improved FT system leads to the conceptualization of the S/88 role as an attached slave I/O processor -it is an I/O handler or pseudo-channel for the S/370.
  • all of the basic communication between the processors must be initiated from the S/88 (because of the design).
  • the S/88 can access all of the S/370 memory and microcode space via EXEC370, while the reverse is not true - the S/370 processor 85 cannot access the S/88 storage at all, even accidentally.
  • the truer picture is of the S/370 as slave to the S/88, but with the internal image of a normal stand-alone S/370 with S/370 1/0. The S/370 dons not 'know' that the S/88 is there.
  • S/370 10 instructions must be able to INITIATE an action, and this facility is provided by the PU-BCU request line 256a, which has a singular meaning: S/370 has a high-priority message waiting for S/88 (usually an I O instruction).
  • the priority nature of this service demand is the reason for the automatic mailbox scheme and the linked-list programming of DMAC channel 0.
  • the DMAC 209 is an integral part of the BCU hardware design. It is initialized and basically controlled by S/88 firmware, and data transfers are paced by the BCU logic which drives the four request REQ input lines 263a-d, one for each channel. In addition, external BCU logic activates the Channel 0 PCL line 257a as each mailbox transfer completes, causing the DMAC 209 to present an interrupt request to the S / 88 processor 62.
  • the initialization and programming of the DMAC 209 is entirely standard and preferably in conformance with the MC68450 Architecture. Briefly:
  • the DMAC 'thinks' the device has 16-bit data, but external logic causes 32-bit transfers.
  • the linked array chaining mode used in CHO (Channel 0 of DMAC 209) implies that a linked-list exists, and it is set up by the ETIO initialization routine. Once CHO is started, it stops only due to an error condition or by encountering the last valid entry in the linked-list. In normal operation, an interrupt to S/88 occurs each time the DMAC 209 completes a mailbox read, and the firmware monitors and replenishes the linked-list in real time; thus the last valid entry of the list is never reached, and CHO runs (idles) continuously.
  • Each DMAC channel is provided with two interrupt vector registers NIV, EIV (Fig. 18), one for normal end-of-operation and one for end forced by a detected error.
  • the present improvement uses all eight vectors, with eight separate ETIO interrupt routines in microcode store 174.
  • the channel 0 normal interrupt has two possible meanings: a PCL-caused 'mailbox received', and the less-common 'channel stopped due to the end of linked-list'. The interrupt handler differentiates these by testing a DMAC status bit.
  • the S/88 firmware also provides four service entries for the EXEC370 application program: initialization, and starting of the three basic data transfers discussed above - data read, data write, and message-Q write.
  • the ETIO-INITIALIZE entry is usually called soon after power-up, but can also be used to re-initialize for error recovery attempts. It resets the BCU hardware and the DMAC 209, then programs the DMAC registers in all four channels with configuration and control values. It also builds the necessary linked-list and starts Channel 0, causing the DMAC 209 to auto-load the first linked-list parameter set and then wait for a request transition from the BCU hardware on line 263a.
  • the other three service entries are called to start DMAC channels 1 (data read), 2 (data write), and 3 (message-Q write).
  • the calling program (EXEC370) provides a pointer to a WORK QUEUE block which has been pre-set with data addresses, count, etc. These routines either start the DMAC209 and BCU hardware immediately, or enqueue the operation if the required DMAC channel is busy. (A separate 'work-pending' queue, shown in Fig. 41 E, is maintained for each of these three channels).
  • a third, small but crucially important, area of S/88 custom firmware is the modification of the S/88 OS (Operating System) to intercept and vector the eight DMAC interrupts to the custom handlers but transparent to the S/88 OS.
  • All of the S/88 firmware for the preferred embodiment is written in MC68020 assembler language, and so cannot properly be termed microcode. It is considered firmware because of the nature of its functions.
  • Each of these instructions is implemented in microcode so as to pass all essential information to EXEC370 in the S/88 via the mailbox mechanism, while maintaining conformance to S/370 Architecture.
  • the existing S/370 microcode exception handler is modified for the 'Adapter Attention' case.
  • the code tests adapter 154 status to determine the cause of the request, and customizes only the Q-not-empty' (which means message received) handling; any other cause returns to existing unmodified code for handling.
  • the defined categories of received messages are:
  • Message types 0001-0006 above are S/370 manual operations for state control, resulting from user input at the (emulated) S/370 System Console. They may also be forced directly by EXEC370 as needed for error recovery or synchronization.
  • Message type 0007 is used to inform the S/370 of asynchronous changes of status of 1/0 devices, such as power-loss, ON/OFF-LINE changes, device-detected errors, etc. It may also be expanded for general-purpose communication from the S/88 to the S/370.
  • Message type 0008 is the vehicle for reporting end-of-I/O operation status to the S/370 - either normal or error end conditions. It will always result in an eventual Program Interrupt and Device Status Table modification in the S/370.
  • Fig. 38 illustrates the microcode design for a preferred embodiment of the present improvement.
  • the code running in the S/370 processing unit (each processing element such as 85) is kept in control store 171 and interprets S/370 instructions when they are executed by PE 85.
  • the microcoded instructions for Start I/O, interrupt handling, operator functions, machine check and initial microprogram load/program load (IMUIPL) are designed specifically to interface with the S/88 microcode as shown in the figure.
  • the interface includes the common hardware facilities of the interface logic 81 including the local store 210, S/370 cache 340 and S/370 real storage space 162 with interrupt capability to the microcode of both processors 85 and 62.
  • the S/370 microcode driver includes CCW convert, interrupt handler, error handler, IMUIPL and synchronizing code interacting with a S/88 application interface (EXEC/370) and the S/88 05.
  • the fault tolerant processor 62 executes all I/O, diagnostics, fault isolation, IPUIML, and synchronization for the system. This system is not viewed as a coprocessor system because S/370 programs are the only programs executing from the users point of view. The system administrator can control the systems attributes through the S/88 fault tolerant operating system. The primary function of the S/88 OS and the application EXEC/370 is 1/0 conversion with a multiple 370 channel appearance. All system error and recovery functions and dynamic resource allocation functions are handed by the S/88 OS. Machine check and operator functions previously handled by the S/370 OS are now passed to the S/88 OS so the functions can be handled in a fault tolerant fashion.
  • Fig. 39 illustrates the execution of a S/370 1/0 command, in this example a start I/O command.
  • the actions taken by the S/370 instruction, S/370 microcode, the coupling hardware (PE85 to PE62), the coupling microcodeo ETIO (executed on PE62) and the S/88 program EXEC 370 are shown briefly, the final step being the execution of the S/370 SIO on the S/88 processor PE62.
  • Fig. 40 is a simplified overview illustrating briefly certain of the components and functions of the improved system in relation to EXEC 370 and the microcode driver used during SIO execution, together with control flow, data flow, signals and hardware/code partitioning.
  • S/370 PE85 microcode and EXEC370 communicate with each other via a "protocol", Fig. 41A.
  • PE 85 microcode sends messages to EXEC370 requesting the execution of functions like 1/0, and EXEC370 sends messages indicating the completion of I/O functions, messages regarding I/O device and channel status changes, and messages requesting PE85 microcode to perform specific S/370 CPU functions.
  • These messages (described in detail later) are transmitted between PE85 microcode and EXEC370 via hardware which includes cache controller 153, adapter 154, BCU 156 and its DMAC 209, etc. This message transmission service is made available to EXEC370 by ETIO.
  • the interface Fig. 41 B between EXEC 370, the S/370 External support software executed by S/88 and the BCU microcode driver (ETIO) running on PE 62 consists of a set of queues and buffers residing in the store 210, one event id, an EXBUSY variable, and a subroutine call sequence.
  • the subroutine CALL interface initiates data transfer operations between S/88 and S/370 and initializes the DMAC 209 and BCU 156 at S/88 reboot time.
  • the queue interface is used to keep track of work items until they can be processed, and the event ID interface (an interrupt to S/88) notifies EXEC370 when work has been added to the queues.
  • Each of the twenty seven Work Que Buffers holds data pertaining to one specific task or service request. Twenty six WQBs are used to service PE85 microcode initiated requests. The remaining WQB (EXWQB) 501-31 is reserved for servicing requests originated by S/88 and sent to PE85 microcode; it will never appear on the freeQ Fig. 23E. Each WQB is addressed by a base address and an offset value stored in DMAC 209.
  • Each WQB, Fig. 41 D contains a 16 byte mail block 505, a 1.6 byte parameter block 506, and a 224 byte device specific work area 507.
  • the mail block 505 contains data passed between EXEC370 and PE85 microcode. Its content is transparent across the ETIO interface.
  • the parameter block 506 contains parameters passed between ETIO and EXEC370, usually with respect to the transferring of data between local store 210 and main store 162.
  • the work area 507 is owned by EXEC370. It contains information about the progress of the requested operation, current S/370 device status, possible user data, type of S/88 device, pointers to other EXEC370 control blocks, error occurrence information, etc.
  • the mail block 505 includes four fields containing S/370 1/0 information passed between PE85 microcode and EXEC370:
  • the parameter block 506 contains six parameters used when data transfer is requested between store 210 and main store 162 by EXEC370.
  • EXEC370 uses queues for maintaining the WOBs.
  • the queue communication area 501-4 is 256 bytes long and resides at offset 400 (hex) in the store 210.
  • Fig. 41 E shows the queues defined between ETIO and EXEC370 for holding pointer entries to WOBs:
  • Fig. 41 E shows the path of WQBs through the queues. All queues are initialized by EXEC370 during S/88 reboot. Empty WQBs are kept on the freeQ. ETIO removes them from the freeQ as needed to fill the link lists 516.
  • the DMAC 209 via the link list 516, places S/370 mailbox entries from mailbox area 188 of storage 162 into the mail block areas of empty WQBs. WQBs on the link list which have been filled are moved to the workQ 511 by ETIO. When ETIO puts one (or more) WQBs on the workQ 511 and EXEC370 is not busy, ETIO notifies the EX370 event ID. EXEC370 removes the WQB from the workQ before it services the request.
  • EXEC370 calls ETIO which initiates the proper BCU156 operation or, if the hardware resource is busy, puts the WQB on the appropriate S/370 Q.
  • Each of the three services send messages to S/370, transfer data to S/370 and transfer data from S/370
  • WQBs are added to one of the S/370 queues by ETIO code while on the EXEC370 thread.
  • the ETIO interrupt routine puts the WQB on the 588 Q 515; and, if EXEC370 is not busy, notifies the EX370 event ID.
  • Fig. 42 illustrates the movement of WQBs through queues together with interfaces between EXEC 370, ETIO, interface hardware 89 and S/370 microcode.
  • 10 interrupt if any
  • EXEC370 returns the WQB to the freeQ.
  • EXEC370 gets its next task by checking first the S88 Q 515 and then the workQ 511. If both are empty, EXEC370 sets an EXBUSY variable to zero and waits for the FX370 event to be notified. EXEC370 sets EXBUSY to 1 when it is notified, before it begins processing.
  • All queues, the EX370 event ID, and the EXBUSY variable reside in the queue comm area 501-4 of store 210 as shown in Fig. 41 F.
  • Each queue is circular in nature as shown in Fig. 41 G, with two index type pointers: a fill index 517 and an empty index 518.
  • the fill index 517 points to the next queue entry to fill, and the empty index 518 points to the next entry to empty. If the empty equals the fill index, the queue is empty. All six queues will never overflow since each has 32 entries and there are only 27 WQBs.
  • Each queue also includes:
  • the hardware communication area contains 1024 bytes.
  • the BCU communication area uses 512 bytes of address space.
  • the link list 516 Fig. 41 H is used by the DMAC209 to bring in mail block items from the mailbox area 188 of store 162.
  • WQBs from the freeQ 510 are used to fill entries in the link list 516.
  • Each link list entry contains ten bytes, and identifies the address of the WQB in store 210 in which to put the data, the byte count of the data to be transferred (16), and the address of the next link entry in the list.
  • the DMAC 209 (channel 0) interrupts S/88 when it comes to a link list entry with a zero next link-dress.
  • the current position of the DMAC 209 (channel 0) in the list is available to the software at all times.
  • ETIO In addition to its interrupt entry points, ETIO has two external callable entry points:
  • EXEC370 calls etio init once per S/88 reboot, while EXEC370 is initializing. The queues have already been initialized and the event ID fields will be valid. PE85 microcode will not be operating yet, however it may be in the process of IML (initial microprogram load).
  • EXEC370 calls etio(wbn) whenever it wishes to have data or messages transferred from/to S/370.
  • the parameter wbn is a two-byte integer Work Queue Buffer Number identifying the WQB containing the service request.
  • Wbn is an index value, ranging from 0 to 27.
  • the service request is identified by the req field in the Parameter block.
  • the subroutine ETIO queues this WQB on the S/3701 Q, S/3702Q or S/3703Q, if the requested I O function cannot be initiated immediately.
  • the ETIO interrupt routine will dequeue the next WQB from the appropriate S/370 Q when the previous operation finishes.
  • PE85 microcode should not be notified (eg by an interrupt) until the mail block entry is in the S/370 message queue area 189 of store 162.
  • EXEC370 and S/370 microcode require a Device Status Table (DST) with an entry for each I/O device in S/370 store 162.
  • DST Device Status Table
  • EXEC370 and S/370 microcode communicate with each other via 16-byte messages (see mail block 505 Fig. 41 D) which are sent back and forth. There is a queue which holds the messages in FIFO order for the receiver on each end. There is also a notification mechanism (PU to BCU, and BCU to PU lines).
  • the 16-bit S/370 opcode field "op" contains a request or response from either EXEC370 or S/370 microcode.
  • the 16-bit Channel Unit Address (CUA) is the operand address of a S/370 1/0 instruction.
  • CAW is a 32-bit content of hex location 48 in S/370 storage 162 when the I/0 instruction was issued and includes the storage key.
  • the 8-byte CCW is addressed by the above CAW.
  • EXEC370 returns an interrupt indication, this field contains the CSW.
  • PE 85 stores the CSW in S/370 hex location 40 when it causes the 1/0 interrupt.
  • the CUA field will be unchanged.
  • the OPERATION message is sent to EXEC370 by S/370 microcode whenever a S/370 instruction is encountered which is to be partially or completely handled by EXEC370.
  • the OPERATION message contains the information described above with respect to the mail block 505 of Fig. 41 D.
  • the EXEC370 messages sent to S/370 microcode include:
  • S/370 microcode maintains a table containing information about the status of each addressable S/370 Device.
  • the major pieces of information are:
  • a CSW Channel Status Word
  • the channel is masked OFF the CPU does not accept the CSW.
  • S/370 Microcode saves the CSW and sets DST (CUA) condition to 01.
  • a subsequent TIO or SIO will result in the saved CSW being stored and the condition code 01 (CSW stored) being placed in the CR.
  • S/370 microcode When S/370 microcode is initialized, it will assume all Devices are not operational. S/88 will send an ONLINE message for each device to be supported. The device is identified by its CUA (Control Unit Address).
  • Figs. 44A-L illustrate microcode sequence flows utilized for the execution of these S/370 I/O instructions.
  • the BCU 156 (and adapter 154) is the primary hardware coupling mechanism for effecting the ultimate S/370 1/0 instruction execution by the S/88 hardware.
  • the DMAC 209 is the main "traffic policeman" for directing the flow of operations and data.
  • Channel 0 of DMAC 209 receives I/O commands from the S/370, channel 1 handles data flow from S/370, channel 2 handles data flow to S/370 and channel 3 sends interrupt (and other) messages to S/370.
  • the local store 210 in BCU 156 forms the communication area between the S/370 and S/88.
  • the local bus 223/247 couples the S/88 processor 62 to the DMAC 209 and to local store 210.
  • the local bus 223/247 couples the DMAC 209 and store 210 to S/370 via speed-up hardware in the BCU 156 and adapter 154.
  • S/370 I/O instructions are dispatched to S/370 microcode routines for handling within the S/370, and a S/88 application program EXEC 370 (together with its related S/88 ETIO microcode) effect the ultimate I/O execution.
  • the adapter 154 and BCU 156 form the hardware connection between the S/370 and S/88 code.
  • the start 1/0 microcode routine has a table DST which keeps track of the status of each device, e.g., is it currently available, did it already issue a SIO, is it busy, has it received an interrupt back. This information is contained in the condition code CC.
  • the channel to be cleared is addressed by bits 16 through 23 of the instruction address.
  • S/370 microcode receives control from dispatch, it begins by checking the channel address. The channel address will be either valid or invalid. If the channel address is invalid, the condition register (CR) is set to 3 and S/370 returns to the next sequential instruction. A channel which is supported by S/370 microcode is considered to have a valid channel address. For channel address valid, S/370 microcode sends a clear channel message to EXEC370. It then goes through all the device status table (DST) entries for this channel. All the condition code fields are set to 3 meaning not available, and any pending interrupt table (PIT) entries found are released to a free pit list. S/370 microcode then sets the condition register to 0 and goes to the next sequential instruction.
  • DST device status table
  • EXEC370 when it receives the clear channel message performs an I/O system reset for all devices on the addressed channel. It then ascertains which devices will be on line and sends a status message to S/370 microcode to redefine the configuration on that channel.
  • S/370 microcode When S/370 microcode receives the status message it modifies the condition code in the device status table for each device addressed to it in the status message.
  • S/370 microcode When S/370 microcode receives control from dispatch, it gets the control unit address CUA from the upper end address of the instruction. Using the control unit address it finds the correct device status table DST entry for this device. It checks the value of the condition code CC. There are three options: (1) CC equals zero or 3, (2) CC equals 2 or CC equals 1 and next condition NC equals 2 and (3) CC equals 2 or CC equals 1.
  • CC equals zero or 3
  • S/370 microcode merely sets the condition register to the value of CC and goes to the next sequential instruction.
  • S/370 sends a clear 1/0 message to EXEC 370. It waits for the acknowledgment and clears any pending interrupt entries associated with the device. It then waits for the interrupt message to be returned by EXEC370. Meanwhile when EXEC370 receives the clear 1/0 message, it performs its selective reset of the addressed device, builds a control status word for the device and returns an interrupt message back to S/370 microcode. When S/370 microcode receives the interrupt message, it generates the PIT entry and fills in the NC and CSW from the message. The pit entry is then connected to the DST entry.
  • CC equals 2 or CC equals 1.
  • the first path is the device is busy or the device has sent a pending interrupt but remains busy. This is the case for the selective reset being issued.
  • the second path is where the device has a pending interrupt but is no longer busy. For both of these paths, CC will be equal to either 2 or 1.
  • S/370 microcode pops the interrupt, puts the CSW in S/370 storage, sets the condition register to 1 and returns to the next sequential instruction.
  • Halt Device (Fig. 44C) -
  • S/370 microcode When S/370 microcode receives control from dispatch for a Halt device instruction it checks the condition code for the addressed device status table entry. There are three options, a condition code equals 0 or 2, condition code equals 1, or condition code equals 3. For the first option, condition code equals 0 or 2, S/370 microcode sends a halt device message to EXEC370. It then zeros the 16 status bits in the S/370 CSW, sets the condition register to 1 and returns to the next sequential instruction. Meanwhile when EXEC370 receives the halt device message, it performs the appropriate function on the addressed device and returns a normal interrupt message.
  • S/370 microcode pops the interrupt from the PIT table, puts a CSW in the proper location in S/370 storage, sets the condition register to equal 1 and goes to the next sequential instruction.
  • CC 1
  • S/370 microcode merely sets the condition register to equal 3 and goes to the next sequential instruction.
  • Halt I/O (Fig. 44C) - At this level of description, the function for halt I/O is identical to the function for halt device.
  • S/370 microcode When S/370 microcode receives control from dispatch for a resume I/O instruction, it checks the condition code for the addressed device status entry. There are two options. CC equals 0, 1 or 2 and CC equals 3. For CC equals 0, 1 or 2. S/370 microcode sends a Resume I/O message to EXEC370, sets the condition code to 2 and sets the condition register to 0 and goes to the next sequential instruction. Meanwhile when EXEC370 receives the resume I/O message, it will look up the control unit address and continue the previously suspended I/O operation. For the second option, CC equals 3 - S/370 microcode merely sets the condition register to 3 and goes to the next sequential instruction.
  • Start I/O (Fig. 44E) -
  • S/370 microcode When S/370 microcode receives control from dispatch for a start I/O instruction, it uses the control unit address to find the device status table entry. It then checks the condition code and there are one of four options. CC equals 0, CC equals 1, CC equals 2 and CC equals 3. For CC equals 0, the device is ready and S/370 microcode sends a start I/O message to EXEC370, sets the CC equal to 2 meaning busy, sets the condition register to 0 meaning accepted, and returns to the next sequential instruction. Meanwhile when EXEC370 receives a start 1/0 message, it uses the control unit address to find the specific device and begins a normal I/O operation on that device.
  • CC For the second option, CC equals 1, S/370 microcode pops the interrupt, puts the CSW into S/370 storage, sets the CSW busy bit "on", sets the condition register equal to 1, and returns to the next sequential instruction.
  • CC For the third option, CC equals 2, S/370 microcode sets the CSW and S/370 storage location 40X to all zeros, turns the CSW busy bit on, sets the condition register equal to 1, and goes to the next sequential instruction.
  • CC For the fourth option, CC equals 3, S/370 microcode merely sets the condition register equal to 3 (meaning device not operational) and goes to the next sequential instruction.
  • S/370 microcode When S/370 microcode receives control from dispatch for a start I/O fast instruction, it checks the condition code for the addressed DST entry. There are two options, CC equals 0, 1, or 2 and CC equals 3. For the first option, CC equals 0, 1 or 2, S/370 microcode sends a start I/O fast message to EXEC370, sets the CC equal to 2, the condition register to 0 and goes to the next sequential instruction. Meanwhile when EXEC 370 receives a start I/O fast message, if it is able it starts the I/O operation; otherwise it returns an interrupt message with a CSW containing a deferred condition code which acts as a normal interrupt when it is received by S/370 microcode. For the second option, condition code equals 3, S/370 microcode merely sets the condition register to 3 and goes to the next sequential instruction.
  • Test I/O (Fig. 44G) -
  • S/370 microcode receives control from dispatch for a test I/O instruction, it checks the condition code.
  • CC 0 or 3
  • CC 1 or CC equals 2.
  • the microcode sets the condition register equal to the CC value and goes to the next sequential instruction.
  • CC 0 or 3
  • the microcode pops the interrupt and puts the CSW in S/370 storage, sets the condition register to 1 meaning CSW stored, and goes to the next sequential instruction.
  • CC the microcode zeros the CSW area (40X) in S/370 storage, sets the CSW busy bit "on”, sets the condition register equal to 1 and goes to the next sequential instruction.
  • S/370 microcode When S/370 microcode receives control from dispatch for a store channel ID instruction, it checks the channel address. There are two options, channel address valid and channel address invalid. For the option channel invalid, the microcode sets the condition register equal to 3 and goes to the next sequential instruction. For the option channel address valid, the microcode sets S/370 storage location, A8 hexadecimal to hexadecimal 20000000. It then sets the condition register to 0 and goes to the next sequential instruction.
  • Test Channel (Fig. 441) - When S/370 microcode receives control from dispatch for a test channel instruction it checks the channel address. Note for this flow there are two major options and three minor options. For the first major option, channel address invalid, the microcode sets the condition register to 3 and goes to the next sequential instruction. For the second option, channel address valid, the microcode further checks all DST entries for this channel. The first minor option occurs if the microcode discovers a DST entry for a specific device with CC equals 1 meaning this device has a pending interrupt. For this case, the microcode sets the condition register to equal 1 and goes to the next sequential instruction.
  • Primary and Secondary Interrupts are S/370 terms.
  • a primary interrupt contains at least the Channel End (CE) status bit in the CSW resulting from an I/O operation.
  • a secondary interrupt is either a second interrupt containing the Device End (DE) for the I/O operation; or it is an asynchronous interrupt initiated by the device requesting service.
  • CE Channel End
  • DE Device End
  • the difference between the I/O masked and the I/O enabled interrupts of Figs. 44J and K is whether the 1/0 is masked. That is, whether the S/370 processor will accept an interrupt coming from the channel or not. If an interrupt is not accepted by the S/370 processor, the channel stacks the interrupt; and it is termed a pending interrupt until such time as the S/370 processor is enabled.
  • an interrupt condition occurs while tile EXEC370 is emulating a specific device operation, it builds a CSW and stores it in a message which it then sends to the S/370 microcode.
  • the microcode When the microcode receives this interrupt message it checks the S/370 mask to find out if the 1/0 is masked or enabled. If the I/O is masked (Fig. 44J) it stacks the interrupt. A description of the stacking interrupt process is set forth below. If S/370 microcode checks the mask and I/O is enabled, (Fig. 44K) the condition code field in the DST entry for the interrupting device is set equal to the next condition (NC) in the interrupt message, the CSW from the message is put into S/370 storage, and the microcode causes an 1/0 interrupt to be performed.
  • S/370 mask If S/370 microcode checks the mask and I/O is enabled, (Fig. 44K) the condition code field in the DST entry for the interrupting device is set equal to the next condition (NC) in the interrupt message, the CSW from the message is put into S/370 storage, and the microcode causes an 1/0 interrupt to be performed.
  • S/370 1/0 Masking Events (Fig. 44L) - If the 1/0 is masked when the EXEC370 sends an interrupt message to S/370 microcode, the interrupt is stacked in a pending interrupt table (PIT) entry. At a subsequent point in time, some S/370 event will occur which results in the enabling of I/O interrupts. This could be due to a load PSW instruction, a set system mask instruction, or any interrupt for which the mask enables I/O. At any point when the PSW system mask is changed in such a way as to enable previously masked I/O, S/370 microcode must check for any interrupts pending for those channels. If none are found, the microcode merely exits to the next sequential instruction. If one is found however, the microcode pops the interrupt off the table, puts the CSW in S/370 storage and performs an I/O interrupt.
  • PIT pending interrupt table
  • the BCU 156 performs the I/O interface function between the S/370 chip set 150 and the I/O subsystem which is comprised of the S/88 PE62 and its associated system and 1/0 components in module 10.
  • the S/370 chip set 150 and the I/O subsystem communicate via the bus adapter 154.
  • the S/370 storage area 162 within the S/88 main storage 16 is sometimes referred to herein as the basic storage module (BSM) 162.
  • BSM basic storage module
  • the BCU 156 includes a 64KB local store 210, a direct memory access controller (DMAC) 209, a 32 bit local address bus 247, a 32 bit local data bus 223 and interface logic 205.
  • DMAC direct memory access controller
  • the DMAC 209 includes four 4 data transfer channels:
  • the DMAC 209 transfers double words (32 bits) between the bus adapter 154 and the local storage 210. It also interrupts the I/O subsystem (S/88 PE62) when I/O data transfers are complete.
  • the local store 210 includes I/O and message data buffers WQBs and link-list data for auto-mailbox loads via DMAC 209.
  • the BCU logic 205 includes a local bus arbitration unit 216 in which the S/88 PE62 and the DMAC 209 contend for access to the local bus, i.e., data bus 223 and address bus 247.
  • the PE62 'Bus Request' line 190 is active whenever the following addresses (see Fig. 41 C) are detected by the address decode and arbitration unit 216:
  • the DMAC Bus Request line 269 is active when it wishes to gain control of the local bus 223,247 for a DMAC sequence (read or write the local storage 210) or a link-list load sequence (read from the local storage).
  • the bus grant line 268 is raised when control of the local bus is given to the DMAC 209 by logic 216; line 191 is raised if control is given to PE62.
  • the BCU logic 205 controls the DMAC 209 transfer timing between the bus adapter 154 and the I/O subsystem and converts up to 4KB I/O transfers into 64 byte block transfers for the bus adapter 154 on the channels 0 and 1.
  • BCU logic 205 detects a 64 byte boundary crossing for any block transfer. If this should occur, the block will be broken into two separate transfers. The BCU 156 will calculate the number of words up to the 64 byte boundary for th first transfer. This will be presented, along with the starting address to the bus adapter 154. The remaining words, along with a new address, will be presented to bus adapter 154 via a subsequent command (BSM read/BSM write). BCU logic 205 also provides a pre-empt of 1/0 data transfers (on a 64 byte boundary) as a high priority message or mailbox read request occurs. A high priority message request and a mailbox request can be handled concurrently in the BCU 156. A 'BSM Read' and "BSM Write" operation can be handled concurrently in the BCU 156.
  • the BCU 156 performs the following four 1/0 operations:
  • the 'Start 1/0 instruction SIO, the 'Channel Address Word' CAW and the first 'Channel Control Word' CCW are stored in predetermined 'mailbox' locations in S/370 storage 162. This information is passed to the local storage 210 via the BCU interface logic 205 and bus adapter 154.
  • the DMAC Channel 0 registers shown in Fig. 18 are used for mailbox read operations. They will be programmed by the S/88 PE62 to operate in a Linked Array Chaining Mode'. The PE62 initializes this mode by setting up a series of 'linked lists' (tables) in the local storage 210, Fig. 41 H. It will then set the first 'top linked list address' into the DMAC Channel O Base Address Register (32 bits) BAR. This address points to the first location in store 210 of the linked list data.
  • the DMAC 'PCL' (Peripheral Control Line) 257a will be programmed by PE62 to cause the DMAC 209 to activate its IRQ interrupt output line 258 whenever the PCL line 257a is activated.
  • the 'PCL' line 257a will be activated following the completion of a mailbox data transfer from main storage 162 to the local storage 210 via adapter buffer 259. The interrupt will inform the S/88 processor PE62 that a mailbox load has just completed.
  • the link list data (Fig. 41 H) consists of the following: the starting storage address of a data block; the storage transfer count; and a link address to the next table entry. The last link address in the table will be zero.
  • the S/88 processor 162 sets the top linked list address in the DMAC Channel 0 base address reg.
  • the S/88 processor PE62 will activate the DMAC 209 by writing a "1" into bit 7 ('START' bit) of its channel 0 channel control register CCR.
  • the DMAC 209 will then read the first linked list into its channel 0 registers as follows:
  • the S/370 PE85 decodes a 'START 1/0' instruction, it places the 'START I/O' command, the Channel Address Word, and the first channel control word in successive 'mailbox' locations which are contained in S/370 memory 162.
  • the starting address of the mailbox (base + queue length) is stored in the base register of the bus adapter 154 at initialization time.
  • the S/370 PE85 issues a 'LD OSCW' control op via the processor bus with bit 11 active. This sets the 'PU to BCU REQUEST' bit on in the control word of the bus adapter 154.
  • OSCW bit 11 causes a 'PU to BCU Request' on the adapter bus (Channel O). If a 'PU to BCU REQ' occurs during an 1/0 data transfer, then the BCU 156 will preempt the I/O transfer on a 64 byte boundary to allow for a mailbox load to take place.
  • the BCU 156 then generates on bus 290 a 'Read Mailbox Select Up' command in the format shown in Fig. 45A where bits 0,1 are the command bits and bits 2-7 are the byte count and stores this in Channel 0 command register 214.
  • the mailbox address bits are stored in register 219 via bus 290 in a format shown in Fig 45B where bit 7 identifies the IOA area in storage 162; bits 24-26, the BCU Channel Number and bits 27-31, the mailbox offset.
  • the BCU 156 activates the COMMAND/STATUS bus 249 and ADDR/DATA bus 250, by filling registers 214 and 219, it raises a 'TAG UP' command on line 262a and waits for data from bus adapter. It does this by sampling 'TAG DOWN' line 262b. 'TAG DOWN' is active as long as data is not ready. As soon as 'TAG DOWN' is de-activated by bus adapter 154 (data ready), the first four bytes of mailbox data is latched in the Channel 0 read buffer 226 via two channel 0 subcycles.
  • the BCU logic 253 then raises 'REQUEST' line 263a on channel 0 of the DMAC 209.
  • the DMAC 209 then raises 'BUS REQUEST' (BR) to line 269 to the LOCAL BUS arbitration circuit 216. If the local bus is not being used by the S/88 processor 62, bus access is granted via bus grant line (BG) line 268 to the DMAC 209.
  • the DMAC 209 then transfers the starting address of the WQB Local Mailbox (in store 210) from MAR to the Address Bus 247, and raises 'ACKO' (DMAC Channel O acknowledge) line 264a.
  • the 'ACKO' signal initiates the transfer of the data from buffer 226, via the data bus 223, to the Local Mailbox portion of the WQB in store 210.
  • the 'DTACK' line 265 is activated to inform the DMAC 209 that the operation is complete.
  • the BCU clock signals (Fig. 25) continue to transfer mailbox data from the buffer 259 to register 226.
  • the BCU 156 performs two adapter bus ('TAG UP'/'TAG DOWN') sequences (16 bits each) for each local storage 210/DMAC 209 sequence (32 bits).
  • the DMAC 209 When the DMAC cycle is complete (DTACK active), the DMAC 209 raises 'Data Transfer Complete' (DTC) line 267 to the BCU logic 253 which then issues another 'REQUEST" to DMAC 209 on line 263a to read the second four bytes from register 226 to the WQB mailbox.
  • the DMAC cycles repeat until the entire mailbox data (16 bytes) has been transferred (4 local bus cycles).
  • the 'PCL' line 257a will then be activated by the BCU logic 253 to the DMAC 209. This causes the 'IRQ' line 258 to be activated from the DMAC 209 to the S/88 processor priority encoder/interrupt logic 212. PE62 will then handle the mailbox request.
  • the DMAC 209 When the DMAC 209 completes its channel 0 register loads from the linked list, it then waits for a signal on Channel 0 'REQ" line 263a from the BCU logic 253 to begin the next mailbox load. Once started, the DMAC Channel 0 remains active indefinitely, with the S/88 processor 62 controlling the circular linked list, and the BCU 156 suspending data transfers by keeping the 'REQ' line 263a inactive. If channel O stops due to an end-of-list' condition, the S/88 processor will receive a termination interrupt and restart Channel 0 when appropriate.
  • All I/O read and write transfers originate from the S/88 processor 62 via adapter bus architected 'BSM READ SELECT UP' and 'BSM WRITE SELECT UP' commands.
  • the S/370 CCW command and starting address (in S/370 memory 162) is derived from the CCW for a 'START I/O'. Data is moved by the S / 88 processor 62 between each I/O device and a local buffer in local storage 210.
  • the local store 210 includes a queue of storage blocks for I/O Write Operations which is managed by the S/88 processor 62. When the queue includes at least one entry, it is ready to kick off an I/O Write operation. The starting address for a selected one of these blocks is stored in the DNA channel 1 registers in the DMAC 209 by the S/88 processor 62 prior to the initiation of a write operation.
  • the DMA Channel 1 registers are reserved for S/370 I/0 write operations (S/370 storage 162 to I/O) via local store 210.
  • the adapter data buffer 259 (64 bytes) is reserved for mailbox read and S/370 1/0 write operations (data transfers from S/370 memory 162 to local storage 210). This buffer is associated with the Channel O adapter bus 249,250.
  • the buffer 260 (64 bytes) is reserved for message write (to S/370) and S/370 1/0 read operations (data transfers from local storage 210 to S/370 memory 162). This buffer is associated with the Channel 1 adapter bus 251,252.
  • the S/88 processor 62 initializes the high order words of DMAC Channel 1 and 2 memory address registers to zero (0). This saves an extra bus cycle when these registers are loaded during operational sequences, since the local storage 210 does not require more than 16 bits of address.
  • the high order data bus bits (31-16) will be loaded into the low order (15-00) part of the channel 1 memory address register.
  • the high order bits (31-16) of the MAR were set to O during initialization.
  • the DMAC 209 responds with a 16 bit port 'DSACK' signal lines 266a, b via the BCU logic 253 to the S/88 processor CPU.
  • the S/88 processor 62 places the BCU data (byte count, storage key, adapter bus priority and customer/IOA space data) and the DMAC channel 1 memory transfer count data on the local address bus 247.
  • Fig. 45E shows the command on address bus wherein bits
  • the byte count, storage key (derived from the CCW), adapter bus priority, and customer/IOA space bits will be placed on the data bus 223 by the S/88 processor 62 in the format shown in Fig. 45F wherein the bit designation is as follows:
  • the DMAC 209 will load the high order word (i.e., byte count) of the data bus into the channel 1 MTC register.
  • the BCU 156 will capture the data bus contents as follows:
  • the DMAC 209 responds with a 'DTACK' signal on line 265 to the BCU logic 253.
  • the BCU logic 253 converts the 'DTACK' signal to a 32 bit port 'DSACK' response on lines 266a, b to the S/88 Processor 62.
  • the transfer byte count, along with the remaining data bus data will be presented to bus adapter 154 during the subsequent 'BSM READ SELECT UP command.
  • the BSM read boundary counter 221 or the BSM read select-up byte counter 220 will be loaded into the Channel O read command register 214.
  • the BSM starting address on bus 223 is stored in the A/D register 219 and the BSM Read Address register 231. It will be sent subsequently to bus adapter 154 for presentation to the S/370 storage 162.
  • the BCU 156 then activates the 'DSACK' lines 266a, d to the S/88 processor 62. At this point, the S/88 processor is released, and is no longer involved with this operation.
  • the BCU 156 places the 'BSM SELECT UP' (Read) command into register 214 via bus 290 and on the command/status bus 249 shown in Fig. 451 wherein bits
  • the field length was previously transferred from register 220 or 221 into register 214.
  • the register 219 places address information on bus 250 in the format shown in Fig. 45J wherein bits
  • the BCU logic 253 then raises TAG UP line 262a to bus adapter 154 in order to latch the command, field length data into adapter command register 124 (Fig. 13) and key address data into register 122.
  • Bus adapter 154 raises TAG DOWN to the BCU logic 253 if data is not valid.
  • the BCU logic 253 waits until TAG DOWN drops.
  • Bus adapter 154 converts the adapter bus BSM SELECT UP command to a processor bus I/O Memory command as shown in Fig. 45K and 45L in which bits on the processor address/data bus 170 represent:
  • the bus adapter data buffer 259 When the addressed data is returned from S/370 memory 162, it is latched in the bus adapter data buffer 259 (Channel O).
  • the bus adapter 154 then de-activates TAG DOWN line 262b on the adapter bus channel 0. This condition alerts the BCU 156 to-"latch two bytes (16 bits) of data, immediately followed by another two bytes in the Channel O Read Buffer 226 (4 bytes) via the clock left and clock right signals.
  • the BCU 156 then activates its 'REQ1' line 263b (DMAC Channel 1 request) to the DMAC 209.
  • the DMAC 209 issues a 'BUS REQ' on line 269 to the BCU local bus arbitration logic 216, in order to perform a local bus cycle.
  • the DMAC 209 When bus grant signal on line 268 is returned from the BCU arbitration logic 216, the DMAC 209 starts a Channel 0 Read Buffer 259 to the local storage 210 operation. It does this by returning ACK1 (DMA Channel 1 acknowledge) on line 264b to the BCU logic 253, and by gating the local storage address in DMAC channel 1 register MAR to the store 210 addressing circuits (not shown) via bus 248, latch 233, address bus 247 and multiplexor 232.
  • the BCU logic 253 uses the ACK1 signal on line 264b and RAM select signal on line 210a to gate the first data (4 bytes) from buffer 226 to the data bus 223 for storage into store 210 at the address specified by the MAR register.
  • DMAC 209 raises DTC (data transfer complete) on line 267.
  • the BCU 156 will decrement the byte count which has been retained in registers 220, MTC; increment channel 1 MAR; and decrement address register 231 for each double word (4 bytes) of data that is received from bus adapter 154, up to 64 bytes. The sequence described above is repeated for each four bytes (up to 64) of the BCU command. If the transfer byte count is greater than sixty-four, then the BCU 156 will present a new BSM starting address to bus adapter 154 via registers 231, 219 in order to fetch the next 64 bytes. The register 231 has been decremented for each four byte transfer as described above and therefore has the appropriate next starting address. The bus adapter 154 buffers 64 bytes of data for each starting address until the entire data transfer (up to 4KB) requested by the command is complete.
  • the BCU 156 will leave the DMAC 209 idle (by not raising REQ) if the bus adapter buffer 259 is empty, and until the next valid data word is received; the state of tag down reflects the availability of valid data in buffer 259.
  • the REQ/ACK cycles continue until the byte count goes to zero at which time the DMAC 209 raises IRQ on line 258 to the S/88 processor 62. This alerts the S/88 processor 62 to read the local storage buffer which contains the data read from S/370 storage 162 for appropriate processing.
  • 1/0 Read Operations (under the control of EXEC370) are kicked off when at least one entry exists in the 1/0 Read queue in store 210.
  • the S/88 processor 62 gains control of the local bus if it is not being used by DMAC 209.
  • the high order data bus bits 31-16 will be loaded into the low order (15-00) bits of the Channel 2 memory address register.
  • the high order bits (31-16) of the MAR were set to zero during initialization.
  • the DMAC 209 responds with a DTACK signal on line 265 which is converted to DSACK signals on lines 266a, b to the S/88 processor 62.
  • the S/88 processor 62 then moves data (up to 4KB) from an 1/0 controller such as 20 or 24 to the local storage 210 via S/88 program control, using the starting address of the selected local storage 1/0 read buffer.
  • the S/88 processor 62 places DMAC Channel 2 memory transfer count selection on address bus 247 in the format shown in Fig. 450 wherein bits:
  • the DMAC 209 will load the (byte count) of the data bus 223 into the channel 2 MTC register.
  • the BCU 156 will capture the data bus contents when the above command appears on the address bus 247.
  • Bits 26-16 are stored into BSM write select up byte counter 222.
  • Bits 13-07 are stored into the high order byte of adapter bus channel 1 A/D register 227.
  • the DMAC responds with a DTACK signal on line 265 to the BCU logic 253.
  • the logic 253 converts the DTACK signal to a 32 bit port DSACK response on lines 266a, b to the S/88 processor 62.
  • the transfer byte count, along with the remaining eta bus data will be presented to bus adapter 154 during the subsequent BSM write select up command.
  • the count in the BSM write boundary counter 224 all but last transfer
  • the BSM write byte counter 222 last transfer
  • the S/88 processor 62 then generates a BSM write select up command on the local address bus 247 in the format shown in Fig. 45Q wherein bits
  • the S/88 processor will also place the BSM starting address on the data bus 223 in the format shown in
  • the BSM starting address on the data bus 223 will be captured by the low order bytes of the Channel 1 A/D register 227 and BSM write address register 228. It will subsequently be sent (as seen below) to bus adapter 154 for presentation to the S/370 storage 162.
  • the BCU 156 then activates the DSACK lines 266a, b (32-bit port) to the S/88 processor 62. At this point, the S/88 processor 62 is released, and is no longer involved with this operation.
  • the BCU logic 253 issues a BSM select up command, gating bits "01 into the high order bits of command register 225 via bus 290 and places the command and field length of register 225 on bus 252 in the format shown in Fig. 45S wherein bits
  • register 227 The contents of register 227 are placed on the address/data bus 251 (in two subcycles) in the format shown in Fig. 45T, wherein bits
  • the command, field length are stored in register 125 of adapter 154.
  • the key / address data is stored in register 123 of adapter 154 via SYNC register 113.
  • the BCU logic 253 activates the REQ2 signal on line 263c to the DMAC channel 2.
  • the DMAC 209 sends the I/O buffer starting address from MAR to store 210 via bus 248, latch 233, bus 247, multiplexor 232 to transfer a double word of data from store 210 to A/D register 227.
  • ACK2 (DMA Channel 2 acknowledge) is raised on line 264c. This causes a Tag Up on line 262a to adapter 154.
  • the adapter 154 then transfers a double word of data from the register 227 to bus adapter buffer 260 in two subcycles via register 113.
  • a write sequence of REQ/ACK signals followed by a Tag Up command is repeated to transfer each double word of data.
  • the BCU 156 decrements the byte count in registers 222,224 and the address in register 228 and MTC of DMAC channel 2 for each double word (32 bits) that is presented to bus adapter 154 up to 64 bytes.
  • the BCU 156 will present a new starting address for the next 64 bytes.
  • the bus adapter buffers 64 bytes of data for each starting address. This sequence will repeat until the byte count in register 222 (4KB max) goes to zero.
  • the BCU 156 will suspend the write sequence until the bus adapter gives a buffer available indication via the Tag Down line 262c.
  • the bus adapter 154 converts the adapter bus BSM Select Up command to a S/370 Processor Bus I/O Memory command in a format shown in Fig. 45U and V on the processor bus 170 and the key/status bus wherein
  • All high priority message transfers originate from the 1/0 subsystem (S/88 processor 62).
  • the DMAC channel 3 will be set up by the S/88 processor 62 to perform the data transfer (16 bytes).
  • the BCU 156 will use the adapter bus channel 1 for data communication (Q Select UP command).
  • the BCU 156 detects a high priority message request when the S/88 processor PE62 performs a DMAC memory transfer count load to register MTC in channel 3. As a result of this, the BCU 156 generates a Q Select Up command to the S/370 PE85 on adapter bus 252 of channel 1. If a S/370 I/O read data transfer (adapter bus channel 1) is in progress when the request is detected, then the BCU 156 waits until the current 64-byte block transfer is complete before honoring the request.
  • PE62 gains control of the local bus 223, 247 if it is not being used by the DMAC 209. PE62 then stores the message data in the local storage 210 via program control. PE62 sets the local buffer message starting address in the DMAC channel 3 memory address register MAR by placing information on the local address bus 247 in the format shown in Fig. 45W wherein bits
  • the starting address of local buffer message data intended for the memory address register is placed on the data bus 223 in the format shown in Fig. 45X wherein bits
  • the high order data bus (Bits 31-16) will be loaded into the low order (bits 15-0) part of the DMAC channel 3 memory address register MAR.
  • the high order bits (31-16) of MAR were set to zero during initialization.
  • the DMAC 209 responds with a DTACK signal on line 265 which is converted to a 16-bit port DSACK signal on lines 266a, b via the BCU logic 253 to the S/88 processor 62.
  • the S/88 processor 62 then places a command on the local address bus 247 in the format shown in Fig. 45Y wherein bits
  • the byte count, storage key and customer/IOA space bits will be placed on the data bus by the S/88 processor 62 in the format shown in Fig. 45Z wherein bits
  • the DMAC 209 will load the high order word (byte count) of the data bus 223 into the channel 3 memory transfer count register MTC.
  • the BCU 156 will capture the data hus contents when this particular command appears on the address bus 247 by storing bits 19-16 into the Q Select Up counter 254 and bits 11-07 into channel 1 A/D register 227.
  • DMAC 209 responds with a DTACK signal to logic 253 which converts it to a 32 bit port DSACK response on lines 266a, b to PE 62.
  • This action alerts the BCU 156 to initiate a high priority message transfer from local storage 210 to the S/370 BSM 162.
  • the transfer byte count, along with the additional data shown in Fig. 45Z are presented to bus adapter 154 during a BCU generated Q Select Up command.
  • the Q Select counter 254 is loaded into bits 4-7 of the channel 1 write command register 225.
  • the BCU 156 places the Q Select Up command in register 225 via bus 290; and the data in register 225 is placed on the adapter bus 252 (channel 1) in the format shown in Fig. 45AA wherein bits
  • the data on buses 252 and 251 is transferred into adapter registers 125 and 123 respectively.
  • the BCU logic 253 then activates the REQ line 263d (DMA Channel 3 request).
  • the DMAC 209 places the I/O buffer starting address (from MAR) on the local bus 247, and raises ACK (DMAC Channel 3 acknowledge) line 264d.
  • the BCU 156 then transfers the first four bytes of data from the addressed I/O buffer in local storage 210 to the adapter buffer 260 in two subcycles via the SYNC register 113. Succeeding four byte blocks are transferred by sequences directed by the Tag Up command to bus adapter 154, and the REQ/ACK lines 263d, 264d to the DMAC.
  • the BCU 156 decrements the byte count for each double word (32 bits) that is presented to bus adapter 154.
  • the bus adapter 154 converts the Q Select Up command to a S/370 processor bus I/O memory command to send the message to area 189 of storage 162; the format of the command is shown in Fig. 45AC wherein PROC BUS 170 bits
  • the processor 85 KEY/STATUS bus has data in the format shown in Fig. 45AD wherein bits
  • the DMAC 209 presents interrupt vectors from the least significant byte of its data bus 248 to the S/88 processor data bus 161 D, bits 23-16 via driver receiver 234 and bits 23-16 of the local data bus 223.
  • the DMAC returns a 16-bit DSACK to PE 62.
  • a Read BCU Status Command can be issued by the S/88 processor 62 in order to read the current status of the BCU 156.
  • the command is placed on the address bus 247 by the S/88 processor 62 in the format shown in Fig. 45AE wherein bits
  • the BCU 156 will place the status shown in Fig. 45AF on the data bus, and return DSACK (32 bit port) on bus 266 PE 62.
  • the bits in Fig. 45AF represent
  • BCU status bit 21 (BSM Read Sync Check) will be reset after it is read by the S/88 processor 62. This bit indicates that the bus adapter 154 and BCU 156 byte counts do not agree when a BSM Read operation terminates; hence an error is detected which requires re-sync.
  • bus adapter 154 will activate Tag Down 262b to indicate that all data has been received.
  • Tag Down 262b will then be deactivated by bus adapter 154, at which time the status indicators will be presented to, and captured by the BCU 156. If Tag Down is not deactivated within 100us, the BCU 156 will activate a cancel line (not shown) to bus adapter 154. This will then cause bus adapter 154 to disconnect itself from the BCU 156.
  • Tag Down 262b is also used by bus adapter 154 to indicate any error that cannot be reported to the BCU 156 via the Command/Status bus 252.
  • a programmed BCU Reset issued by PE 62 performs the same function as a Power on Reset to the BCU 156. It can be issued at any time in order to clear the BCU of any abnormal conditions. However, a local bus cycle (007EXXXX decode) must be recognized by the hardware for this command to execute.
  • the command is placed on the local address bus 247 by the S/88 processor in the format shown in Fig. 45AG wherein bits
  • the data bus contents will be ignored by the BCU 156.
  • the BCU 156 will return DSACK (32 bit port) on lines 266a, b to the S/88 processor 62.
  • Emulation of S/370 DASD on S/88 will be described by way of example to illustrate a preferred manner in which S/370 1/0 programs can be executed by S/88 processors and 1/0 devices.
  • the S/370 is referred to as the Object system, and the S/88 as the Target system.
  • DASD Direct Access Storage Device
  • S/370 code running in the S/370 processor is referred to as object system software. The discussion is divided into four parts:
  • DASD physical media is partitioned into cylinders and the cylinders into tracks.
  • the number of each and their capacity varies for different DASD types and models.
  • Each cylinder is program addressable by a two byte cylinder number (CC), and individual tracks within a cylinder are accessed by separate read/write heads each of which is addressable by a two byte head number (HH).
  • CCHH head number
  • the physical location of a track is given by its cylinder and head number and is therefore specified by the four byte track address (CCHH).
  • Each track contains a home address, a track descriptor (record 0), and one or more data records.
  • the size of each record is programmable; and when the home address and record sizes are written on a track, that track is said to be formatted. All tracks are formatted from their track index to the following track index.
  • Fig. 46A illustrates one such track.
  • the basic unit of information recorded on the physical media is a data byte consisting of eight bits.
  • a group of data bytes makes up an area, and the device separates these areas by writing gaps between them.
  • Each record consists of two (count, data) or three (count, key, data) areas, while the home address is made up of only one area.
  • the three areas making up an object system record are: count, key (optional), and data.
  • the first area on each track is the home address. It contains the following fields:
  • Record O is always the first record following the home address area.
  • the record O CCHH field defines the alternate track if the track has been flagged as defective.
  • the Key Length is normallyzero for record O. Record 0 may be followed by one or more data records.
  • the key area is optional, and if present may contain from 1 to 255 bytes.
  • the number of a record is determined when a Format Write CCW command writes the count, key and data areas. After the record has been formatted, the user data areas may be read and/or rewritten (using other CCW commands) without destroying adjacent records on the track. If a record is reformatted, those following it on the same track are destroyed.
  • DASD (Fig. 46B) is presented to S/88 microcode in the form of files which contain 4096 byte blocks of data sequentially numbered from one.
  • the emulation mechanism maps object system format and function into a useable target system format and function combination.
  • the physical parameters of different DASD types and models in the object system vary.
  • the DASD type and model number along with the various parameters are kept in the first data block, INFO, of the target system file, Fig. 46C.
  • the balance of the file contains the emulated object track data Fig. 46C.
  • the data for each track is maintained in an integral number of data blocks.
  • the number of target system data blocks required for each track is a parameter kept in the first data block.
  • Each emulated track contains a directory of the records currently existing on that track, a directory header, and the user data (key, data) for each record.
  • the directory is used to locate the data for a specific record, perform search on record or key operations, access the last record on the track, and handle track overflow.
  • Object system data is treated in the emulation environment in one of three ways: Maintained, Retained implicitly, or Not kept.
  • the CCHH (track address) passed object system software is used to calculate the location of the emulated track in the target system DASD file. It is kept in the track header described below, but is not proliferated throughout the count and home address areas of the emulated track. The home address is not kept as an explicit area.
  • the record number (R), also passed by object system software, is maintained implicitly and does not appear as explicit data.
  • the balance of the object system data [F (logical record track overflow), KL, and DL] is maintained in the track directory, Fig. 46E.
  • a directory entry contains F, KL, and DL, as well as a pointer p to the user data (KEY and DATA) for each record. R is maintained implicitly as the directory entry number.
  • Fig. 46E shows rhe header, directory and user data makeup as well as the mapping of an emulated track into the target system 4KB blocks. Pointers p0-p2 point to the beginning addresses (within 4KB blocks) of user data records 0-2.
  • Figs. 46F-K inclusive represent data transferred by the object system software during read and write operations.
  • ops CCW operations
  • CCHH values of Fig. 46F are calculated and/or checked but nothing is written to the emulated track.
  • CCW ops involving count always orient the head to the next record in track Fig. 46H.
  • CCW ops involving key and data the location and size of the user data is found in the directory, Fig. 461.
  • CCW ops involving count, key and data orient the read/write head to the next record in track, Fig. 46J.
  • processing begins with next directory entry and continues to the last valid directory entry, Fig. 46K.
  • Fig. 10 conceptually illustrates S/88 virtual storage 106 and physical storage 16 and the allocation of S/370 physical storage areas 162-164 for S/370 processors 21, 23 and 25, 27 and 29, 31;
  • Fig. 47 which illustrates diagrammatically the method of capturing one S/370 storage area from the S.88 physical storage 16;
  • Storage 16 is divided into 4KB pages and a plurality of storage map entries (mme), one for each 4KB page, are contained in mme arrays (Fig. 48A) which together map the entire storage 16.
  • the entries corresponding to pages not assigned for use are tied together in a "freelist" (i.e., the storage allocation queue) by including in each entry (Fig. 48B) the physical page numbers (pointers) of the previous and next entries in the list.
  • a software pointer in the S/88 operating system always points to the beginning-of the freelist.
  • Physical storage pages are assigned to various processes from the beginning of this freelist and pages returned to the freelist are preferably placed at the beginning of the freelist.
  • the "previous and next" page numbers and the software pointer to the beginning of the freelist are updated appropriately.
  • Service routines in the application program EXEC370, described below, provide the functions for stealing S/370 storage areas from the S/88 operating system.
  • Fig. 48A is a simple overview of the software structure set up by the S/88 operating system (S/88 OS) to maintain a process's virtual address space.
  • the software structure includes the following elements:
  • Storage Map Array of Fig. 47 conceptually illustrates all of the mme arrays arranged in sequential address order. Mme's are usually threaded onto one of three lists:
  • the mme data structure used by the storage management unit 105, contains the three list pointers shown in Fig. 48B wherein:
  • the "next" and “previous” mme fields are used to create the chained lists (used, reclaim, freelist).
  • each mmep array (Fig. 48C) is a list of 128 pointers, each of which is a virtual address of an mme array.
  • the first n pointers are an ordered list of all the mme arrays.
  • Each of these pointers comprise the 16 most significant (high order) bits of a physical address, called a physical page number (ppn), and are used as a pointer to a specific mme.
  • the seven high order bits of the ppn select the mme array, and the nine low order bits of the ppn select the mme within the array.
  • the twelve low order bits of the physical address are an offset into the real (physical) page of storage 16.
  • a memory map information (mem map info) structure (Fig. 48D) is used to keep track of memory used for maps, wherein:
  • the active page table entries are used to keep track of virtual storage. There is one apte for each 4KB page of virtual storage in all virtual storage spaces in the system.
  • the apte structure (Fig. 48E) indicates the owner(s) of the virtual space, the virtual address of the page, and the real memory address of the disk address if paged out.
  • the apte structure includes:
  • Each apte is twelve bytes long, and 256 entries are contained in each active page table (apt) block (Fig. 48F).
  • the relative position of apte's within a block has no significance. All unused apte' s are threaded onto a free aptep list. If additional apt's are needed and the list is null, a new apt block is allocated in the wired heap; and the entire 256 apte's are threaded onto the free aptep list.
  • the apt trailer (Fig. 48G) is used for shared program regions, it is allocated in the wired system heap, and pointed to by an EITE (executable image table entry) or an apte. There will be four trailers per program (one per region). Trailers allow the system to find all PTWs which point to a page when removing it.
  • the apt trailer structure includes:
  • the process table entry (pte) (Fig. 48H) contains the information needed to manage a process; it contains information about the process's virtual address space.
  • Each page table entry includes:
  • the process map block structure (Fig. 481) is used to map a process's virtual space into real memory space and includes:
  • the process management table (Fig. 48J) contains information used by the scheduler, including a list of pointers ptep to all the processes in the system, the number of pages available in the system and the number of pages committed.
  • the physical table word (ptw) of Fig. 48K includes:
  • the System/88 includes a startup procedure that powers on the system and boots program and data modules which are included in a startup file.
  • the programmable read only store (prom) 181 (Fig. 12) runs diagnostics and self tests on both the System/88 and System/370 components. At the completion of those tasks, the PROM 181 reads a utility program that loads the S/88 operating system from a master disk (not shown).
  • the module start up code initializes all configured devices and disks and sets the internal clock from the system calendar clock.
  • This file contains commands that the operating system executes as part of the procedure for starting up a module.
  • This procedure includes functions of:
  • the module file supplies sufficient.data to bring up a new system and can be modified by the customer to suit his requirements.
  • certain statements are inserted into the module startup code command file. For example, if we assume the configuration of Fig. 10 with three S/370 processors 21, 23 and 25, 27 and 29, 31 and three S/370 storage areas 162, 163 and 164 for said processors, the following statements are inserted into the module startup code command file:
  • Each Start S/370 command causes a software routine, to be executed to "steal" a block of real storage space from storage 16 for the particular S/370 processor #1, #2 or #3. Then the appropriate S/370 operating system will be IPLed into the "stolen” real storage space.
  • the functions of the software routine are to capture storage areas from S/88 storage and to "replace" those areas when appropriate. Five subroutines are used to perform these functions:
  • FIRST MME pertains to the first mme in the string which is to he unthreaded, and base ppn contains its ppn (physical page number).
  • LAST MME pertains to the last mme in the string. If the FIRST MME is at the head of the free list (its previous mme field equals zero) the free list pointer is set equal to the next_mme field of the LAST MME; thus the mme following the LAST MME is now at the head of the free list. Otherwise the next mme field of the mme previous to the FIRST MME is set equal to the next_mme field of the LAST NME. If there are mme's following the LAST MME (its next mme field not zero), the previous mme field of the mme following the LAST MME is set equal to the prev mme field of the FIRST MME.
  • a configuration array is built in S/88 kernel storage which contains the base ppn and n blks for each S/370 processor.
  • n blks means the number of contiguous mega-bytes of storage. It is equal to the number of stolen (unthreaded) mme's divided by 256.
  • registers 404, 405 are coupled directly to the S/88 processor data bus 161 D (rather than to the BCU local data bus 223).
  • Decode logic 280 of logic 216 decodes the above virtual address to block AS from the S/88 hardware and to return DSACK to processor 62.
  • Registers 404, 405 are enabled via the STCI select line 458 from logic 253. Bits 27-20 of the STCI word form the STCI "base" address and bits 23-20 form the S/370 storage "size” value. Bits 19-0 are zeros.
  • a first method involves modification of the S/88 operating system kernel by inserting the S/370 interrupt handler code into the S/88 operating system first level interrupt handler so that it is assembled as part of that object module.
  • the table of interrupt vectors are contained in the interrupt handler assembly source, and the vectors used by S/370 are modified in the source to point to the S/370 interrupt handler code.
  • the above first method circumvents all of this.
  • S/370 interrupt vectors to point to the S/370 interrupt routine, we have avoided all of the normal interrupt processing performed by the S/88 operating system and do not have to identify S/370 via the device file. This is really a software uncoupling, since code has been modified instead of hardware.
  • This first method is the quickest and least expensive method to achieve the desired interrupt function.
  • this method is susceptible to additional maintenance for each subsequent release of the S/88 operating system. It requires at least a kernel bind; and, if the interrupt handler has been changed, the S/370 interrupt code must be reinserted and the interrupt handler reassembled.
  • a second method involves modification of operating system interrupt vectors after system boot; and it is this method which is implied for use with the description of the hardware interrupt mechanism of Fig. 20.
  • This second method requires the placing of the S/370 interrupt code into the S/88 operating system virtual address space (in the preferred embodiment, just below 007EOOOO) and the modification of the appropriate interrupt vectors in the operating system kernel interrupt handler. This work is done by the S/370 initialization routine after the operating system has initialized (at the same time that the S/370 initialization routine "steals" storage). Since the initializing routine is modifying the S/88 operating system kernel storage area, it has to "gain freedom" in the manner set forth to "steal” storage in the above description.
  • This second method does not require the maintenance modification of each newly released S/88 operating system kernel. However, S/370 interupts are nonfunctional until after S/88 operating system is up and running.
  • a third method involves hardware presentation of interrupt vector content; and this is a preferred alternative because no change in the S/88 operating system kernel is required, i.e., no change is made in the vector table.
  • This third method requires the placing of the S/370 interrupt routine into the S/88 operating system virtual address space and/or the BCU Local Storage as a known read-only store (ROS) address.
  • the interrupt routine address(s) must be made available to the S/370 hardware, preferably in ROS. The following scenario is given to illustrate the method:
  • This third method enables uncoupling during board sync for the purpose of syncing S/370 hardware, etc. However, this method requires more hardware.
  • This trap 13 interrupt routine does nothing more than "call” the program issuing the trap at the location immediately following the trap instruction. Since the trap interrupt routine is in supervisor state, the subject program will change to supervisor state. To regain application program state, the application program modifies the interrupt stack return address and returns from the trap 13 "call” to the trap 13 interrupt code which exits from the interrupt using the modified interrupt stack address. This method involves the addition of an interrupt routine to the S/88 OS.
  • a second method eliminates the modification of the subject OS.
  • a special register (not shown) is defined in the BCU control storage address space which, when written to by the application program, causes a new BCU interrupt using the third method for implementing interrupts given above.
  • the application interrupt routine is made resident in BCU read-only storage (not shown) and functions the same as the trap 13 code.
  • the Gain Freedom routine previously described functions exactly the same except that it writes to the BCU special register instead of issuing a trap 13 instruction.
  • the "stealing of memory” does not require reassembly of any S/88 source code or binding of the S/88 OS kernel.
  • the address of the head of the free list is available to the application program.
  • each S/88 processing element 60, 62 (Fig. 8) of unit 21 drives the A bus 42 and the B bus 44.
  • This concurrent operation is in contrast to I/O units in the processor module 9 which implement an error check prior to driving the bus structure.
  • the processing unit 21 operates in this manner because timing therein is sufficiently important that any delay in operation is undesirable for system throughput.
  • An error noted by the checking logic during the time the processing unit is driving the bus structure, causes the unit to drive both an A Bus Error signal and a B Bus Error signal onto the X bus 46 during the next phase of the system clock.
  • the failing central processing unit (e.g., 21) drives a level 1 maintenance interrupt, onto the X bus 46, which the partner central processing unit (e.g., 23) receives.
  • the failing unit goes off-line, becoming incapable of driving further signals onto the bus structure 30, except in response to interrogation from the partner central processing unit.
  • This automatic off-line operation ensures that any read or write cycle is aborted, whether to the memory unit 16, 18 or to a peripheral device through a control unit during which an error was detected in either the address or the data on the A bus or B bus. Further, any data transfer during that same operating cycle is repeated using only the partner central processing unit.
  • the comparator 12f compares the input data which the processing section 12a receives from the A bus 42 with the input data which the processing section 12b receives on the B bus 44. It also compares the function, address and data signals (including parity) which the processing section 12a applies to transceivers with corresponding signals which the processing section 12b produces. Timing and control signals of section 12a are compared with corresponding signals from section 12b. This comparison of internal control signals checks internal operations of the processing elements 60, 62 and facilitates prompt detection of faults and is useful in diagnosis and maintenance of the processor unit.
  • the comparator produces a Compare Error signal which is applied to the control stage 86.
  • the error can be the result of a data-in error, a data-out error, a function error or an address error. It can be, also, either a cycle error or a control error due to differing timing or control signals.
  • the detection of an error by the parity-checking circuits produces a Parity Error signal which is applied to the control stage 86.
  • the control stage 86 responds to the Compare Invalid signal and the Parity Invalid signal to produce, on the next clock phase (N + 1), a Processor Error signal.
  • control stage 86 produces the Processor Error signal only if no Bus Error signals are produced with the next timing phase.
  • a Bus Error signal indicates a fault condition in the bus structure 30 and hence identifies that the invalid comparison of input data was the result of a fault in the A Bus or B Bus portion of the bus structure 30 and not in either processing section 12a or 12b.
  • One function of the Processor Error signal is to disable logic circuits and thereby essentially halt all operation in the processing section 12 of unit 21.
  • the A Bus Error signal and the B Bus Error signal are applied to the X Bus 46 to signal all units in the module 9 to ignore information placed on the bus during the immediately preceding phase, e.g., to ignore the CPU Bus transfer.
  • a Level One Interrupt signal is applied to the X Bus 46 to notify the partner processing unit 23 that some unit in the module has detected a fault-producing error.
  • the stage 86 still in response to the fault signal, terminates the assertive bus master status. This action is accompanied by the termination of the Bus Error signals.
  • the processing section 12 switches out of the Master state, it disables all the bus drivers in the transceivers 12e.
  • the S/370 transceiver 13 drivers are also disabled via common control 75 whenever those of transceivers 12e are disabled.
  • transceivers 12e, via control stage 86, and transceivers 13 are also disabled.
  • processing units 21, 23 can drive the bus structure only when in the Master state, as required to produce the Bus Enable signal that is applied to the drivers.
  • the Processor Error signal promptly, i.e. at the end of the next timing phase, turns off the master status.
  • the S/88 processing section of partner unit 23 continues operating essentially without interruption.
  • the partner processing unit 23 repeats the data transfer.
  • the partner unit reads in the repeated data which the memory applies to the bus structure in a subsequent timing phase.
  • the partner processing unit 23 responds to the Level One interrupt which is a low priority interrupt, to initiate a diagnostic routine.
  • the cause of the Processor Error appears to be a transient phenomenon, i.e., the diagnostic routine does not identify or locate any faulty or erroneous condition, the processing unit 21 can be restored to operation without maintenance.
  • the occurrence of the transient failure is recorded, and if repeated an arbitrarily determined number of times the processing unit is electrically removed from service or operation without further diagnosis.
  • Each processing section 12 of the units 21, 23 includes logic circuits, typically in the processor status and control stage 86 to bring the two partner units into lock-step synchronization.
  • the section 12 attain lock-step synchronization with the transition to Master status.
  • Each section 12 must be in the Master state in order for it to drive signals onto the bus structure.
  • the initializing sequence stored in each PROM 181 typically includes instructions for bringing the partnered sections into synchronization and to ensure that neither processing section is in the Master state initially, i.e., upon being turned on.
  • the processing sections 12 of the units 21, 23 are not in synchronization initially in the initializing sequence and one unit attains the Master state during a multi-phase cycle prior to the other.
  • the one unit obtaining Master status controls the further initilizing operation of the other unit to bring it into the Master state at a selected time.
  • the processing section 12 of unit 21 When the processing section 12 of unit 21 is initialized, it negates an internal Error Check signal, and thereby prevents a Parity Invalid signal or a Compare Invalid signal from producing a Processor Hold Signal. Instead, the section 12 executes a test routine, typically stored in the PROM 181, which exercises all conditions that can produce a Processor Error signal. As each potentially faulty condition is created, the processing section tests to see whether the corresponding fault reporting signal is indeed produced. The absence of the Error Check signal thus inhibits the processing unit from attaining Master state, with the result that faults produced during this logic exercising routine do not stop the processing unit and are not reported to the bus structure 30. The test routine in the PROM 181 asserts the Error Check signal and enables the processor to assume the Master State only upon successful completion of this checking routine.
  • the S/370 processing units typically have hardware provided for initialization and service processor functions via a "back door" access to the various components and logic in each chip. Since these are well known, they will be described only briefly.
  • the manner in which this code goes about synchronization is to transfer a storage-mapped copy of the register set within a primary (or master) partner processing unit such as 21 (one that is operating properly) to the register set within a secondary (or slave) partner processing unit such as 23 (one that is not yet operating properly).
  • the module 9 is comprised of one or more simplexed S/370 processing units such as 21 or pairs of partner S/370 processing units such as 21, 23. S/88 simplexed or partner units such as 12 or 12, 14 may be included in the module for executing only S/88 programs.
  • Each S/370 processing unit includes a pair of S/370 processor elements such as 85, 87 and a pair of S/88 processor elements such as 62, 64 as shown in Fig. 7; and the pairs of processing elements are operated in lock step as a single logical processing unit.
  • the partner units form a redundant design operated in lock step with each other to provide a fully fault tolerant, self checking logical processing unit.
  • Each of the S/370 processor elements 85, 87 of a pair is in part a S/370 chip set such as 150 (Fig. 11).
  • the S/370 chip sets and their associated hardware are mounted on a S/88 style board such as 101 (Fig. 9A) for coupling with S/88 bus structure 30; and they are coupled to respective S/88 processing elements via interface logic circuits 89 and 91 (Fig. 8).
  • the S/370 chip set pair and their associated hardware in one processing unit such as 21 will be referred to as a S/370 entity; and their corresponding S/88 processing elements such as 60, 62 and associated hardware will be referred to as a S/88 entity.
  • the S/370 entities execute S/370 application programs and call upon the S/88 entities to perform the S/370 1.0 operations as required utilizing the S/88 I/O devices and programs such that neither the S/88 nor the S/370 operating system is aware of the other.
  • One of the more unique and significant features of the S/88-S/370 processing units is the self- determined synchronization of any processing unit such as 21 by a currently-processing partner 23.
  • the S/88 entity of each unit has the capability and the responsibility for the synchronization of a new or error producing partner. When a S/88 entity of a unit assumes this responsibility, it is referred to as the master. Its partner, which undergoes synchronization, is referred to as the "slave.”
  • the S/88 hardware/firmware structure determines when synchronization is required and who synchronizes whom.
  • the interconnected S/88-S/370 hardware/firmware utilizes this same intelligence to follow the lead of the S/88 in synchronization decisions. That is, anytime the S/88 determines that a S/88 (slave) entity requires synchronization with its partner (master), that synchronization is permitted to progress to a suitable point after the S/88 slave entity has been "kicked-off"; then the execution is diverted to the corresponding S/370 entity.
  • the S/370 entities are synchronized by the S/88 PEs executing code from PROM 181 to extract the S/370 Master state and restoring that state to both S/370 partners.
  • Either one of the partner pair can assume the master or the slave role in the synchronization of processing units, whether the requirement is invoked by an Initial Power On, the appearance of a new partner or a recovery from an error condition that caused two existing partners to lose synchronization (each case forcing a Maintenance Interrupt).
  • the S/88 slave entity recognizes its status and depends on the S/88 master entity for synchronization.
  • the S/88 master and slave entities assume their respective roles as a result of their respective states at the time the maintenance interrupt occurs.
  • the S/88 entities of all processing units detect and process the interrupt with each assuming it is a slave until a defaulted master is established. That master then kicks off any holding slave in lock-step, each resuming the pre-empted environment of the master (upon returning from the interrupt).
  • the S/88 entities will uncouple the processors from the rest of the logic, use those processors to emulate the S/370 SP function to establish an identical pre-empted state within the S/370 partner-pair, then will re-establish the normal execution environment and permit the S/370 partner-pair to begin execution in lock-step.
  • the S/88 entity has appropriate hardware facilities for establishing synchronization.
  • the S/370 processing section has sufficient hardware and software assists to permit a slave entity to be initialized to the exact same state as the master entity. This includes such features as read/ write status registers, readable mode registers, clearable caches, stoppable clocks and count rings, etc.
  • the S/370 processing entity provides two methods for accessing the various registers and caches for synchronization.
  • One is the normal, user-programmed read/ write method using registers 560, 561 (Fig. 49) which couple the BCU local data bus 223 to channels 0, 1 of adapter 154.
  • the other is a serial "back-door" Integrated Support Facility (ISF)/Universal Support Interface (USI) (540, 541) approach.
  • ISF/USI serial interface/protocol
  • the synchronization mechanism of the S/88 entities can access any and all facilities associated with the S/370 entity.
  • both methods are employed.
  • the normal path is used where it exists, and the USI path is used for the rest.
  • this part of the synchronization and initialization process (i.e., for the S/370 entities) must be transparent to the S/88 operating system which is not aware of the presence of, or the connection to, a S/370 entity.
  • This transparency is achieved in a manner generally similar to that described above with respect to S/370 I/0 operations. That is, the address decode logic 280 described with respect to Fig. 20, senses an address 007EXXXX each time data is to be transferred between the S/88 processor 62 and the logic of Fig. 49. When this address is decoded by logic 280, it couples the S/88 processor bus 161A, 1610 to the local BCU address and data buses 247, 223 via circuits 217, 218 as described earlier.
  • Register address decode logic 562 decodes the low order bits of the address on bus 247 to select one of the logic circuits 549, 550 or registers 560, 561 for data transfer with processor 62.
  • interrupts on lines 562, 563 are directed to the S/88 interrupt logic 212 of Fig. 20 via OR circuit 292a.
  • the interrupt request signal is activated on line 562 when data has been received in logic 549 from one of the S/370 chips for transfer to processor 62.
  • An interrupt request signal on line 563 notifies the processor 62 of the completion of a data transfer from logic 550 to a S/370 chip.
  • An interrupt request on line 562 notifies the processor 62 that data has been received by logic 549 from a S/370 chip for transfer to processors 62.
  • the interrupt requests are held on lines 562 and 563 until an IACK signal appears on lines 258d and 258c respectively.
  • Vector numbers for these interrupts are derived from logic 564, 565 when energized by IACK signals 258d and 258e respectively from Fig. 20. The vector numbers are used by the processing element 62 to access the respective interrupt handler routines.
  • the S/370 integrated support facility (ISF) 540 Fig. 49 represents a "backdoor" entry to the logic on chipset 150.
  • the ISF consists of a 5 line support bus 541 which connects to the Unit Support Interfaces (USls) integrated on chips 85 and 151-154. A portion of the USI 542 on chip 85 is shown in Fig. 49.
  • the support bus 541 represents a serial interface with the following 5 lines:
  • the ADDR_MODE line 545 signals the serial transfer (shift) of either address bits (up level) or data bits (down level) on the BIT OUT/BIT IN lines 543, 544.
  • the BIT OUT and BIT IN lines 543, 544 are the interconnection between shift registers such as 548 inside a chip and external shift registers in logic 549, 550.
  • the number of bits shifted between an internal register 548 and one of the two external registers 549, 550 is determined by the number of pulses applied to the shift gate line 546.
  • the SET PULSE is used to synchronize chip internal activities based on the address or data pattern just shifted into the chip.
  • SET PULSE is activated after shifting is finished to signal the availability of the information on the chip side e.g., in register 548. This means that activities based on the information can be initiated as of this moment.
  • a start function is assigned to a specific address pattern. This address is shifted into the registers such as 548 of each chip. When all address bits have been transferred, the address decode 551 in one of the chips detects its address. The SET PULSE follows the address transfer. The address decode and the SET PULSE form a chip internal start pulse at the output of gate 552.
  • the chip specific part of a USI contains controls and data chains as derived from the specific chip design. To retain the current status of storing elements not affected by a shift operation, the functional clocks must be stopped prior to the initiation of any USI activities. USI accesses requiring clock stop as prerequisite are defined as 'static'. Dynamic accesses or functions are those operations which can be executed while the chips are in operation.
  • the SET PULSE is used to synchronize functions to the chip internal timing.
  • the functions are decoded from the address pattern or data pattern in the SERDES register, additionally gated by the ADDR MODE line (address or data mode):
  • the five-wire serial bus 541 of the ISF that provides a 'back door' access to the various addressable entities within the S/370 chip set 150, is coupled to the Unit Support Interface (USI) of each chip e.g., USI 542 of chip 85.
  • the USI 542 provides an 8-bit Address Register 566 and an 8-bit Serializer/Deserializer (SERDES) 548.
  • SERDES Serializer/Deserializer
  • the USI Address Register 566 receives the address of the chip and the address of the target entity within the chip while the SERDES 548 is the actual send/receive mechanism.
  • the USI also provides synchronization logic for the shift-in/shift out mechanism.
  • Each chip within the S/370 chip set 150 is assigned a 4 bit (high order) ISF/USI address, for example PE85, cache controller 153, clock 152, adapter 154, floating point coprocessor 151, and STCI 155 being assigned the hexadecimal values of 2, 4, 6, 8, and A and B respectively.
  • the low order 4 bits of the ISF/USI Address define the internal chip entity (e.g., register, function or chain), addressed by the low order 4 bits.
  • the communications scheme is comprised of Shift Chains (also referred to as Function Chains) that, in turn, are comprised of fields that identify the command, the source chip, the destination chip, the data and the target entity within the chips.
  • Shift Chains also referred to as Function Chains
  • the shift chains are as follows:
  • the Command Field of the Function Chain may contain a Write/Control Command (E61) or a Read/Sense Command (F61).
  • E61 Write/Control Command
  • F61 Read/Sense Command
  • An example of a Function Chain is as follows:
  • the approaches to establish synchronization described herein use S/88 program code stored in the PROM 181.
  • the code makes determinations associated with each of the above four situations and sets flags accordingly.
  • the synchronization routines then use those flags to control the code pathing in order to perform the appropriate synchronization and/or initialization.
  • a couple of examples are:
  • the other implementation defined herein is microcode only, permitting the handling of the S/370 synchronization by the S/88 processor entities in emulation of a S/370 Service Processor. This technique may be used where performance and 'time-to-ready' is not critical.
  • a simplexed Processing Unit 21 is Powered On (Hardware Implementation)
  • This situation can be caused by either of two conditions:
  • the S/370 entity of the unit 21 merely follows the lead of the S/88 entity. This is accomplished by the S/88 processor 62, executing code resident in the S/88 PROM 181, completing normal Self Test then determining if this is an Initial Power On or a Power Fail Recovery. If it is a Power On, it continues with normal Initialization; then, assuming it is the DMPU, attempts to issue a SYNC signal. The signal is trapped by the S/370 logic which forces a Level 6 Interrupt to the S/88 processor 62. The Interrupt 6 will be vectored to the S/370 Synchronization microcode in the S/88 PROM 181 (Fig. 19A) (which is mapped into the S/88 address space).
  • the S/370 PE 85 has executed its own STIR then suspended execution at its Sync Point.
  • the S/370 clock 152 also, has initialized itself.
  • the S/88 Level 6 Interrupt Service Subroutine (ISS) (i.e., the S/370 Synchronization microcode) uses the ISF/USI interface of Fig. 44 to emulate the S/370 Service Processor. This SP Emulator will issue Function Strings to invoke the IML function of the S/370 control store 171, though no actual code transfer occurs (the microcode is in the S/88 PROM 181).
  • the next step of the IML Emulation is to broadcast the SYNC to the S/370 entity (the processors 85 and 87) causing the processing unit 21 to step off into execution.
  • the final step of the ISS is to Return-from- Interrupt, causing the processing unit to begin execution of the IPLed state.
  • an emulated service processor 'IPL Button Pushed' Function String will be sent to the S/370 processing unit to perform the IPL function, loading S/370 main storage from disk.
  • the final step of IPL is, then, to pass control to the address specified by location 0.
  • the S/88 entity of the unit 21 executes its Self Test and Initialization Routine (STIR) then will determine if this is an Initial Power On (IPO) or a Power Fail Recovery (PFR). If this is an IPO, the code determines that the unit 21 is a simplexed entity and proceeds with loading the Operating System and executing its start-up' routine.
  • TSR Self Test and Initialization Routine
  • IP Initial Power On
  • PFR Power Fail Recovery
  • the synchronization function becomes a 'dummy' operation as there is no associated partner to be synchronized.
  • Duplexed Processing Units 21, 23 are Powered On -Hardware Implementation
  • This situation can be caused by either or two conditions:
  • each processing unit 21, 23 executes a portion of its Self Test and Initialization Routine (STIR) then attempts to determine whether or not the contents of its associated storage 16 have been compromised (Power Fail state). If so, it will fall back to the normal Power On STIR path. If not, it will attempt to determine if it has a partner or co-resident processing unit that may be the DMPU or whether or not it is the DMPU. If it is, it will assume the DMPU responsibility and attempt to synchronize any other processing units. If it is not the DMPU, it will proceed to the Sync Point and await SYNC.
  • TSR Self Test and Initialization Routine
  • Each S/370 entity merely follows the lead of the S/88 entity.
  • the S/88 entity executing code resident in its PROM 181, completes normal Self Test then determines if this is a Power On or a Power Fail Recovery. If it is a Power On, it continues with normal Initialization; then proceeds to the Sync Point. If this is a Power Fail Recovery, the cache is examined to determine whether or not it is valid. If it is, it may have to update its partner's memory, should that partner's cache be found invalid. If it's own cache is invalid, it must depend upon its partner to update it with valid cache contents. If neither partner can assure valid memory they must, as a pair, continue with normal Power On and Initialization. As the S/88 entities of the processing unit pair approach the Sync Point, each S/88 entity determines whether or not it must assume the DMPU responsibility. If it finds that it is the DMPU, it attempts to issue the SYNC.
  • the sync signal is trapped by the S/370 logic and forces a Level 6 Interrupt to the S/88 entity.
  • the interrupt will be vectored to the S/370 Synchronization microcode in the PROM 181 (which is mapped into the S/88 address space).
  • the S/370 entity e.g., processing elements 85. 87
  • the S/370 entity has executed its own STIR then suspended execution at its Sync Point. If this is a Power Fail Recovery.
  • the S/370 entity goes through a process similar to the S/88 entity process of determining how far back into the Initialization routine it must go in order to assure memory integrity and synchronization.
  • the S/370 clock 152 has initialized itself.
  • S/88 processors achieve synchronization by one of the S/88 pair of processors of the unit 23 issuing a SYNC OUT signal on line 570, Fig. 50. If the partner unit has been initialized and self-tested and is determined to be not BROKEN, it has a signal level on the BROKEN line 571 which is inverted by circuit 572 to gate the SYNC OUT signal through AND INVERT gate 573.
  • the SYNC signal was applied to the SYNC IN line 580 of the drive (d) S/88 processor of a unit 14 via line 577 and inverter 574. It is also applied to the SYNC IN line 575 of the checking S/88 processor of unit 12 via the C bus and inverter 576 to initiate the "kick-off" of all four S/88 processors of units 12, 14 in lock-step.
  • the output 577 of circuit 573 is disconnected from the SYNC IN lines 580 and 575 to prevent kick-off of the S/88 processors. Instead it is connected via line 581 to set a flip-flop 582 in the BCU 156 of the partner unit 21 Fig. 49. It also sets a corresponding flip-flop in the paired BCU (not shown) in the unit 21.
  • the following description will address only one S/370 and associated hardware in unit 21, but it will be appreciated that both S/370 entities are operating in a similar fashion.
  • the flip-flop 582 applies a level 6 interrupt signal to the S/88 processor 62 via line 583, OR circuits 292a and 292 (see Fig. 20), interrupt logic 293 and lines IPO-2. This action is referred to as "trapping" of the S/88 SYNC signal by the S/370.
  • the S/88 processor 62 initiates an interrupt acknowledge cycle in response to a SYNC signal on line 583.
  • the function code and priority level signals from processors 62 are decoded in logic 281, a local BCU bus request is made on line 190 via output 283 of decode logic 281, and gate 291, line 287 and OR circuit 284.
  • processor 62 on line 191 When a bus cycle is granted to processor 62 on line 191, it (together with signals on SYNC line 583, AS line 270 and decode line 283) enables AND gate 294-4 to apply a signal to IACK line 258f. This signal is applied to the vector bit logic 584 (Fig. 49) to apply an appropriate vector number to the S/88 processor 62 via BCU local bus 223, driver-receiver 218 and processor bus 161 D. The signal on line 258f also resets the flip-flop 582.
  • the S/88 processor 62 executes a read cycle to obtain the vector number which is then used by the processor 62 to access the first instruction of an interrupt routine for S/370 synchronization.
  • the last instruction of the synchronization routine generates a SYNC command which applies a SYNC signal to line 586 (Fig. 50).
  • This signal is applied to the SYNC lines 580 and 575 to "kick-off" the S/88 (as well as the S/370) processors of partner units 21, 23 in lock-step.
  • an emulated SP 'IML Button Pushed' Function String will be sent to the S/370 entities in units 21, 23. Rather than performing the entire IML function of DASD accesses, etc., this IML will bypass the I/O processes and load from S/88 Main Storage.
  • the EXEC 370 code will already have fetched the IPL code from DASD and placed it in S/88 Main Storage, awaiting the IPL. The final step of IPL is, then, to pass control to the address specified by location 0.
  • a Maintenance Interrupt invokes the S/88 PROM 181 code.
  • This code synchronizes the S/88 entity of the unit 21, then calls the S/370 STIR, also resident in PROM 181.
  • the S/370 STIR determines that, this being a PFR, the necessary facilities are available and proceeds to synchronize and initialize the S/370 entity or unit 21.
  • a level 6 Interrupt will be posted to the S/88 entity of the current unit 21 upon the insertion of the new board. While the new processing unit is running its STIR, the current processing unit will recognize the Level 6 Interrupt. The Level 6 will go about the process of archiving the pre-empted task environment, determining if the new processing unit is online; and, when it is, returning from the interrupt. As a function of the Return-from-Interrupt, the two units will step off into lockstepped synchronization, resuming the pre-empted task.
  • a Maintenance Interrupt invokes the S/88 PROM 181 code.
  • This code resynchronizes the S/88 entity of the unit 21, then calls the S/370 STIR, also resident in PROM 181.
  • the S/370 STIR determines that, this being similar to a PFR, the necessary facilities are available and proceeds to synchronize and initialize the S/370 entity of unit
  • the failing processing unit will be forced into its STIR while the normally-performing processing unit will be interrupted by a forced Level 6 Interrupt.
  • the Level 6 Interrupt Service Subroutine will go about the process of archiving the pre-empted task environment, determining if the new processing unit is online; and, when it is, returning from the interrupt. As a function of the Return-from-Interrupt, the two units will step off into lockstepped synchronization, resuming the pre-empted task. Should the failing processing unit fail to exit its STIR correctly (e.g., once or a selected number of tries), the normally-performing processing unit will, after an appropriate time, set BROKEN to the S/88 portion of the failing processing unit and its various status reporting facilities.
  • a Maintenance Interrupt invokes the S/88 PROM 181 code.
  • This code resynchronizes the S/88 entity of the unit 21, then calls the S/370 STIR, also resident in PROM 181.
  • the S/370 STIR determines that, this being similar to a PFR, the necessary facilities are available and proceeds to synchronize and initialize the S/370 entity of unit 21.
  • Another compare failure will result in the same action being repeated. After a pre-determined number of iterations, the board will be put off-line permanently and a failure reported.
  • hardware fault-tolerance is shown to have at least three features. There is instantaneous, electrical isolation of a failing field replaceable unit without the propagation of data errors to another element of the system. Dynamic reconfiguration code is provided to remove or add components as required or when the components fail. The capability to remove power from and to apply power to a subsystem or field replaceable unit without the loss of the system is provided - i.e., hot plug capability. The user perceives no loss of function or performance.
  • Processors such as S/370 processors, alien to the subsystems of the patent, may be attached to said subsystems in a manner similar to that shown in the present application relative to the S/88.
  • processors of the subsystems can be uncoupled to permit their use as I/O controllers for the parasitic, attached alien processors.
  • the cache 340 may be the exclusive storage for some valid 1/0 data (rather than storage 162 storing all valid I/O data) as is true in typical S/370 cache systems today.
  • I/O data transfers may take place directly between a S/88 1/0 device such as disk controller 20 and the S/370 storage 162 for more efficient operation.
  • the BCU 156 must still be used for transferring S/370 1/0 commands to the S/88.
  • System 370 storage addresses associated with the commands must be changed to S/88 physical addresses by EXEC370 code while the commands are being converted to S/88 commands.
  • one method is to first flush the section of cache, related to the 1/0 operation, to storage 162 prior to performing the I/O operation.
  • the section of cache related to the I/O operation is invalidated prior to performing the I/O operation.
  • the.function may be performed in the I/O device controller(s) by routines similar to those used by EXEC370 within the S/88 processor 62.
  • Data conversion may also be performed by the EXEC370 application calling conversion routines in the S/88 OS such as ASCII to EBCDIC conversion.
  • Fig. 52 illustrates the data flow for an alternative embodiment in which both of a pair of directly coupled processors are uncoupled from their associated hardware, preferably in a manner generally similar to that described with respect to the S/88 processor 62 of the preferred embodiment to transfer command( and/or data between the processors in a manner transparent to their operating systems.
  • Two processors 640, 641 are coupled to each other via the processor buses 642, 643, driver receiver circuits 644, 645 and a common local storage unit 646.
  • the processors 640 and 641 may have the same or different architectures and the same or different operating systems.
  • Each processor 640 and 641 may have its own hardware (not shown) including main storage and I/O devices for normal processing of programs under control of the respective operating systems. Neither operating system is aware of the existence of or coupling to the processor associated with the other operating system.
  • processor 640 of this alternative embodiment When processor 640 of this alternative embodiment is controlled, however, by an application program to send commands and/or data to the processor 641, it preferably puts a predetermined address on the processor address bus 647 which is decoded by logic 648 to cause circuits 644 to couple bus 642 to local store 646 via local bus 652 for command and data transfer from the processor 640 to the store 646. Decoding of the address also uncouples the processor 640 from its associated hardware to render the transfer transparent to the operating system of processor 640.
  • Uncoupling control logic 649 interrupts the processor 641 when I/O commands and/or data intended for processor 641 have been transferred into the local store 646.
  • the processor 641 (via its application program interrupt handler) is uncoupled from its hardware and reads in the commands and/or data from store 646 into its main storage (not shown) in a manner transparent to its operating system. If the commands and/or data require conversion, the processor 641 utilizes the emulation microcode in the store 650 to perform the required conversion. The processor 641 then processes the converted commands under control of its operating system.
  • the "uncoupling" of the processors 640 and 641 may permit the continuous transfer of a substantial segment of commands and/or data to and from the local store 646 before “recoupling" of each processor to its hardware is permitted. In this manner, fast and efficient data transfers will be achieved.
  • Commands and/or data may be transferred in the opposite direction from processor 641 to processor 640 in a similar manner.
  • the commands and/or data may be converted where required by emulation microcode located in store 651; and the converted commands may be processed in processor 640 under control of its operating system.
  • This alternative embodiment differs in one significant respect from the preferred embodiment; i.e. the processor "initiating" the data transfer is uncoupled from its hardware to send data to the "receiving" processor.
  • This requires the additional function of transferring control to an application program similar to EXEC370/ETIO of the preferred embodiment when an I/O function (transfer commands and/or data to another processor) is to be performed.
  • the means for effecting the transfer of control for certain I/O operations from an operating system to an application program will depend upon the characteristics of the system.
  • the S/370 executes a Start I/O instruction which is processed by the operating system in a normal fashion without "uncoupling" the S/370 processor from its associated hardware.
  • a selected invalid OP CODE may be used instead of a Start I/O instruction.
  • Hardware or microcode decode of the selected invalid OP Code transfers control to a special application program which "uncouples" the S/370 from its hardware for information transfer with processor 641 via storage 646.
  • processor 640 may be controlled to write into only one specific section of store 646; and processor 641 is controlled to only read from said one section. Processor 641 is permitted to write only into a second section of store 646 and processor 640 is permitted to only read from said second section. Processors 640 and 641 are inhibited from writing into the second and one sections respectively.
  • the uncoupling and interrupt mechanisms are operated transparent to the operating systems of both processors 640 and 641 as described with respect to the S/88 processor 62 of the preferred embodiment.
  • the emulation functions can be performed by application programs (rather than by microcode in local storage) in the manner described with respect to EXEC370 in the preferred embodiment.
  • Polling techniques could be used rather than the interrupt mechanism to transfer data between the processors 640, 641; however, such techniques would be inefficient.
  • processor 640 and 641 can perform I/O operations for the other processor, either processor can acquire certain of the I/O environment characteristics of the other.
  • one application in one processor may communicate to a like or different application in a second processor without using the services of the operating system in either processing system.
  • Alien is used to define apparatus which is not known to an operating system because it is not defined in the operating system configuration tables; and therefore the operating system has no device driver for the apparatus and cannot control the apparatus.
  • a special application program running on the operating system is aware of the apparatus and can exercise certain control over the apparatus.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Software Systems (AREA)
  • Quality & Reliability (AREA)
  • Mathematical Physics (AREA)
  • Hardware Redundancy (AREA)
  • Multi Processors (AREA)
  • Memory System (AREA)
  • Techniques For Improving Reliability Of Storages (AREA)
  • Memory System Of A Hierarchy Structure (AREA)
EP90305308A 1989-05-17 1990-05-16 Mémoire principale physiquement unique, partagée par deux ou plusieurs processeurs exécutant leurs systèmes opérationnels respectifs Expired - Lifetime EP0398695B1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US353113 1989-05-17
US07/353,113 US5144692A (en) 1989-05-17 1989-05-17 System for controlling access by first system to portion of main memory dedicated exclusively to second system to facilitate input/output processing via first system

Publications (3)

Publication Number Publication Date
EP0398695A2 true EP0398695A2 (fr) 1990-11-22
EP0398695A3 EP0398695A3 (fr) 1994-02-02
EP0398695B1 EP0398695B1 (fr) 1998-09-02

Family

ID=23387812

Family Applications (1)

Application Number Title Priority Date Filing Date
EP90305308A Expired - Lifetime EP0398695B1 (fr) 1989-05-17 1990-05-16 Mémoire principale physiquement unique, partagée par deux ou plusieurs processeurs exécutant leurs systèmes opérationnels respectifs

Country Status (9)

Country Link
US (2) US5144692A (fr)
EP (1) EP0398695B1 (fr)
JP (1) JP2618072B2 (fr)
AT (1) ATE170643T1 (fr)
BR (1) BR9002304A (fr)
CA (1) CA2009548C (fr)
DE (1) DE69032607T2 (fr)
PT (1) PT94055A (fr)
SG (1) SG45151A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2322209A (en) * 1996-12-23 1998-08-19 Ibm Multi-tasking computer system has shared address space among multiple virtual address spaces
WO2000063778A1 (fr) * 1999-04-16 2000-10-26 Maxnetwall Computers Oy Dispositif et procede d'amelioration de la securite des informations
US9477516B1 (en) 2015-03-19 2016-10-25 Google Inc. Concurrent in-memory data publication and storage system
CN113580399A (zh) * 2021-07-01 2021-11-02 唐山晶琢科技有限公司 一体化的多线切割机罗拉轴支架

Families Citing this family (74)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5838946A (en) * 1990-04-14 1998-11-17 Sun Microsystems, Inc. Method and apparatus for accomplishing processor read of selected information through a cache memory
AU628264B2 (en) * 1990-08-14 1992-09-10 Oracle International Corporation Methods and apparatus for providing a client interface to an object-oriented invocation of an application
US5479640A (en) * 1990-08-31 1995-12-26 International Business Machines Corporation Memory access system including a memory controller with memory redrive circuitry
JPH0833818B2 (ja) * 1990-09-04 1996-03-29 インターナショナル・ビジネス・マシーンズ・コーポレイション コンピュータ・システムの操作方法
JP2945498B2 (ja) * 1991-04-12 1999-09-06 富士通株式会社 システム間通信方式
US5590288A (en) * 1991-07-30 1996-12-31 Restaurant Technology, Inc. Distributed data processing system and method utilizing peripheral device polling and layered communication software
JPH05233570A (ja) * 1991-12-26 1993-09-10 Internatl Business Mach Corp <Ibm> 異オペレーティング・システム間分散データ処理システム
US6286013B1 (en) * 1993-04-01 2001-09-04 Microsoft Corporation Method and system for providing a common name space for long and short file names in an operating system
US5504904A (en) * 1994-02-23 1996-04-02 International Business Machines Corporation Personal computer having operating system definition file for configuring computer system
US5666523A (en) * 1994-06-30 1997-09-09 Microsoft Corporation Method and system for distributing asynchronous input from a system input queue to reduce context switches
US5557783A (en) * 1994-11-04 1996-09-17 Canon Information Systems, Inc. Arbitration device for arbitrating access requests from first and second processors having different first and second clocks
US5630045A (en) * 1994-12-06 1997-05-13 International Business Machines Corporation Device and method for fault tolerant dual fetch and store
US5640526A (en) * 1994-12-21 1997-06-17 International Business Machines Corporation Superscaler instruction pipeline having boundary indentification logic for variable length instructions
US5742829A (en) * 1995-03-10 1998-04-21 Microsoft Corporation Automatic software installation on heterogeneous networked client computer systems
US6466962B2 (en) 1995-06-07 2002-10-15 International Business Machines Corporation System and method for supporting real-time computing within general purpose operating systems
US5632013A (en) * 1995-06-07 1997-05-20 International Business Machines Corporation Memory and system for recovery/restoration of data using a memory controller
US5812775A (en) * 1995-07-12 1998-09-22 3Com Corporation Method and apparatus for internetworking buffer management
US5825774A (en) * 1995-07-12 1998-10-20 3Com Corporation Packet characterization using code vectors
US5796944A (en) * 1995-07-12 1998-08-18 3Com Corporation Apparatus and method for processing data frames in an internetworking device
US5651002A (en) * 1995-07-12 1997-07-22 3Com Corporation Internetworking device with enhanced packet header translation and memory
US5748633A (en) * 1995-07-12 1998-05-05 3Com Corporation Method and apparatus for the concurrent reception and transmission of packets in a communications internetworking device
WO1997004552A1 (fr) 1995-07-19 1997-02-06 Fujitsu Network Communications, Inc. Transmission point-multipoint par l'intermediaire de sous-files d'attente
EP0873611A1 (fr) 1995-09-14 1998-10-28 Fujitsu Network Communications, Inc. Commande de flux commande par emetteur pour attribution de tampons dans des reseaux mta longue distance
US5754788A (en) * 1995-12-28 1998-05-19 Attachmate Corporation Method and system for reconfiguring a communications stack
WO1997026737A1 (fr) 1996-01-16 1997-07-24 Fujitsu Limited Dispositif a multidestination fiable et souple destine aux reseaux mta
US5845094A (en) * 1996-06-11 1998-12-01 Data General Corporation Device access controller and remote support facility for installation of cabling in a multiprocessor system
US6279098B1 (en) * 1996-12-16 2001-08-21 Unisys Corporation Method of and apparatus for serial dynamic system partitioning
US5922057A (en) * 1997-01-10 1999-07-13 Lsi Logic Corporation Method for multiprocessor system of controlling a dynamically expandable shared queue in which ownership of a queue entry by a processor is indicated by a semaphore
US5966547A (en) * 1997-01-10 1999-10-12 Lsi Logic Corporation System for fast posting to shared queues in multi-processor environments utilizing interrupt state checking
US6341301B1 (en) 1997-01-10 2002-01-22 Lsi Logic Corporation Exclusive multiple queue handling using a common processing algorithm
US5926833A (en) * 1997-03-24 1999-07-20 International Business Machines Corporation Method and system allowing direct data access to a shared data storage subsystem by heterogeneous computing systems
US6523105B1 (en) * 1997-04-16 2003-02-18 Sony Corporation Recording medium control device and method
US6772419B1 (en) * 1997-09-12 2004-08-03 Hitachi, Ltd. Multi OS configuration system having an interrupt process program executes independently of operation of the multi OS
US6275984B1 (en) * 1998-11-20 2001-08-14 Sega Of America, Inc. System and method for delaying indirect register offset resolution
US6418505B1 (en) * 1998-12-17 2002-07-09 Ncr Corporation Accessing beyond memory address range of commodity operating system using enhanced operating system adjunct processor interfaced to appear as RAM disk
US6397382B1 (en) * 1999-05-12 2002-05-28 Wind River Systems, Inc. Dynamic software code instrumentation with cache disabling feature
US6298437B1 (en) * 1999-05-25 2001-10-02 Sun Microsystems, Inc. Method for vectoring pread/pwrite system calls
US6735765B1 (en) * 1999-12-07 2004-05-11 Storage Technology Corporation Sharing data between operating systems
US6574753B1 (en) 2000-01-10 2003-06-03 Emc Corporation Peer link fault isolation
US6842811B2 (en) * 2000-02-24 2005-01-11 Pts Corporation Methods and apparatus for scalable array processor interrupt detection and response
JP2001244952A (ja) * 2000-02-29 2001-09-07 Sony Corp 通信制御装置
US7234144B2 (en) * 2002-01-04 2007-06-19 Microsoft Corporation Methods and system for managing computational resources of a coprocessor in a computing system
US7228207B2 (en) * 2002-02-28 2007-06-05 Sabre Inc. Methods and systems for routing mobile vehicles
US7302548B1 (en) 2002-06-18 2007-11-27 Cisco Technology, Inc. System and method for communicating in a multi-processor environment
US7013362B1 (en) 2003-02-21 2006-03-14 Sun Microsystems, Inc. Systems and methods for addressing memory
US7747660B1 (en) * 2003-03-24 2010-06-29 Symantec Operating Corporation Method and system of providing access to a virtual storage device
US8612992B2 (en) * 2003-04-09 2013-12-17 Jaluna Sa Operating systems
EP1467282B1 (fr) * 2003-04-09 2008-10-01 Jaluna SA Systèmes d'exploitation
EP1503286B1 (fr) * 2003-07-30 2014-09-03 Jaluna SA Réseau de plusieurs systèmes d'exploitation
JP2007509387A (ja) * 2003-09-30 2007-04-12 ジャルナ エスアー オペレーティングシステム
US20050091467A1 (en) * 2003-10-22 2005-04-28 Robotham Robert E. Method and apparatus for accessing data segments having arbitrary alignment with the memory structure in which they are stored
US7857701B2 (en) * 2004-03-12 2010-12-28 Microsoft Corporation Silent sign-in for offline games
US7587537B1 (en) 2007-11-30 2009-09-08 Altera Corporation Serializer-deserializer circuits formed from input-output circuit registers
US7873782B2 (en) 2004-11-05 2011-01-18 Data Robotics, Inc. Filesystem-aware block storage system, apparatus, and method
CA2590875C (fr) 2004-11-05 2011-09-13 Data Robotics Incorporated Temoin d'etat de systeme de memoire et procede
US20060185687A1 (en) * 2004-12-22 2006-08-24 Philip Morris Usa Inc. Filter cigarette and method of making filter cigarette for an electrical smoking system
US7545272B2 (en) 2005-02-08 2009-06-09 Therasense, Inc. RF tag on test strips, test strip vials and boxes
JP5139658B2 (ja) * 2006-09-21 2013-02-06 株式会社ニューフレアテクノロジー 描画データ処理制御装置
JP4857201B2 (ja) * 2007-06-20 2012-01-18 キヤノン株式会社 情報処理装置
WO2010093356A1 (fr) * 2009-02-11 2010-08-19 Stec, Inc. Module dram sauvegardé à mémoire flash
US20100205349A1 (en) * 2009-02-11 2010-08-12 Stec, Inc. Segmented-memory flash backed dram module
US8566639B2 (en) * 2009-02-11 2013-10-22 Stec, Inc. Flash backed DRAM module with state of health and/or status information accessible through a configuration data bus
US7990797B2 (en) * 2009-02-11 2011-08-02 Stec, Inc. State of health monitored flash backed dram module
US7983107B2 (en) * 2009-02-11 2011-07-19 Stec, Inc. Flash backed DRAM module with a selectable number of flash chips
US8977831B2 (en) * 2009-02-11 2015-03-10 Stec, Inc. Flash backed DRAM module storing parameter information of the DRAM module in the flash
US7830732B2 (en) * 2009-02-11 2010-11-09 Stec, Inc. Staged-backup flash backed dram module
US8169839B2 (en) * 2009-02-11 2012-05-01 Stec, Inc. Flash backed DRAM module including logic for isolating the DRAM
US9043279B1 (en) * 2009-08-31 2015-05-26 Netapp, Inc. Class based storage allocation method and system
US8819447B2 (en) * 2010-03-10 2014-08-26 Sprint Communications Company L.P. Secure storage of protected data in a wireless communication device
US9754634B2 (en) 2011-11-23 2017-09-05 Smart Modular Technologies, Inc. Memory management system with power source and method of manufacture thereof
US9405578B2 (en) * 2014-07-11 2016-08-02 Accenture Global Services Limited Intelligent application back stack management
JP7153435B2 (ja) * 2017-10-12 2022-10-14 ラピスセミコンダクタ株式会社 不揮発性メモリのデータ書換方法及び半導体装置
CN112154408A (zh) * 2018-04-12 2020-12-29 美光科技公司 重放受保护存储器块命令队列
KR102485487B1 (ko) * 2018-07-18 2023-01-06 에스케이하이닉스 주식회사 반도체장치

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4315310A (en) * 1979-09-28 1982-02-09 Intel Corporation Input/output data processing system
EP0171859A2 (fr) * 1984-08-17 1986-02-19 Koninklijke Philips Electronics N.V. Dispositif de commande d'accès en mémoire pour réaliser des régions protégées dans une mémoire et mémoire munie d'un tel dispositif de commande d'accès
EP0230353A2 (fr) * 1986-01-17 1987-07-29 International Business Machines Corporation Gestion d'un coprocesseur dans un système de traitement de données à mémoire virtuelle et machine virtuelle

Family Cites Families (41)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4004277A (en) * 1974-05-29 1977-01-18 Gavril Bruce D Switching system for non-symmetrical sharing of computer peripheral equipment
US4228496A (en) * 1976-09-07 1980-10-14 Tandem Computers Incorporated Multiprocessor system
US4099234A (en) * 1976-11-15 1978-07-04 Honeywell Information Systems Inc. Input/output processing system utilizing locked processors
JPS547252A (en) * 1977-06-20 1979-01-19 Hitachi Ltd Program control system
US4315321A (en) * 1978-06-16 1982-02-09 The Kardios Systems Corporation Method and apparatus for enhancing the capabilities of a computing system
US4244019A (en) * 1978-06-29 1981-01-06 Amdahl Corporation Data processing system including a program-executing secondary system controlling a program-executing primary system
US4316244A (en) * 1978-11-08 1982-02-16 Data General Corporation Memory apparatus for digital computer system
US4245344A (en) * 1979-04-02 1981-01-13 Rockwell International Corporation Processing system with dual buses
US4325116A (en) * 1979-08-21 1982-04-13 International Business Machines Corporation Parallel storage access by multiprocessors
US4354225A (en) * 1979-10-11 1982-10-12 Nanodata Computer Corporation Intelligent main store for data processing systems
JPS5833972B2 (ja) * 1979-11-12 1983-07-23 富士通株式会社 計算機システム間通信方式
US4400775A (en) * 1980-02-28 1983-08-23 Tokyo Shibaura Denki Kabushiki Kaisha Shared system for shared information at main memory level in computer complex
US4368514A (en) * 1980-04-25 1983-01-11 Timeplex, Inc. Multi-processor system
US4412281A (en) * 1980-07-11 1983-10-25 Raytheon Company Distributed signal processing system
AU551032B2 (en) * 1981-03-31 1986-04-17 British Telecommunications Public Limited Company Safety arrangement in computer control system
US4533996A (en) * 1982-02-23 1985-08-06 International Business Machines Corporation Peripheral systems accommodation of guest operating systems
ATE25779T1 (de) * 1981-10-01 1987-03-15 Stratus Computer Inc Digitale datenverarbeitungsanlage mit zuverlaessigkeits-bus-protokoll.
US4597084A (en) * 1981-10-01 1986-06-24 Stratus Computer, Inc. Computer memory apparatus
US4453215A (en) * 1981-10-01 1984-06-05 Stratus Computer, Inc. Central processing apparatus for fault-tolerant computing
JPS58102380A (ja) * 1981-12-11 1983-06-17 Hitachi Ltd 仮想記憶管理方法
JPS5955565A (ja) * 1982-09-24 1984-03-30 Fujitsu Ltd マルチフア−ムウエア方式
JPS5999576A (ja) * 1982-11-30 1984-06-08 Toshiba Corp 画像情報記憶検索システム装置
US4679166A (en) * 1983-01-17 1987-07-07 Tandy Corporation Co-processor combination
US4591975A (en) * 1983-07-18 1986-05-27 Data General Corporation Data processing system having dual processors
US4564903A (en) * 1983-10-05 1986-01-14 International Business Machines Corporation Partitioned multiprocessor programming system
US4727480A (en) * 1984-07-09 1988-02-23 Wang Laboratories, Inc. Emulation of a data processing system
US4660130A (en) * 1984-07-24 1987-04-21 Texas Instruments Incorporated Method for managing virtual memory to separate active and stable memory blocks
US4754394A (en) * 1984-10-24 1988-06-28 International Business Machines Corporation Multiprocessing system having dynamically allocated local/global storage and including interleaving transformation circuit for transforming real addresses to corresponding absolute address of the storage
US4674038A (en) * 1984-12-28 1987-06-16 International Business Machines Corporation Recovery of guest virtual machines after failure of a host real machine
US4799145A (en) * 1985-04-03 1989-01-17 Honeywell Bull Inc. Facility for passing data used by one operating system to a replacement operating system
US4722048A (en) * 1985-04-03 1988-01-26 Honeywell Bull Inc. Microcomputer system with independent operating systems
US4868738A (en) * 1985-08-15 1989-09-19 Lanier Business Products, Inc. Operating system independent virtual memory computer system
JPS6243703A (ja) * 1985-08-21 1987-02-25 Fanuc Ltd 数値制御システム
US4747040A (en) * 1985-10-09 1988-05-24 American Telephone & Telegraph Company Dual operating system computer
JPS62120565A (ja) * 1985-11-20 1987-06-01 Nec Corp 主記憶領域の割付け制御方式
US4920481A (en) * 1986-04-28 1990-04-24 Xerox Corporation Emulation with display update trapping
US4797810A (en) * 1986-06-26 1989-01-10 Texas Instruments Incorporated Incremental, multi-area, generational, copying garbage collector for use in a virtual address space
US4816990A (en) * 1986-11-05 1989-03-28 Stratus Computer, Inc. Method and apparatus for fault-tolerant computer system having expandable processor section
US5093913A (en) * 1986-12-22 1992-03-03 At&T Laboratories Multiprocessor memory management system with the flexible features of a tightly-coupled system in a non-shared memory system
US4967353A (en) * 1987-02-25 1990-10-30 International Business Machines Corporation System for periodically reallocating page frames in memory based upon non-usage within a time period or after being allocated
US5027271A (en) * 1987-12-21 1991-06-25 Bull Hn Information Systems Inc. Apparatus and method for alterable resource partitioning enforcement in a data processing system having central processing units using different operating systems

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4315310A (en) * 1979-09-28 1982-02-09 Intel Corporation Input/output data processing system
EP0171859A2 (fr) * 1984-08-17 1986-02-19 Koninklijke Philips Electronics N.V. Dispositif de commande d'accès en mémoire pour réaliser des régions protégées dans une mémoire et mémoire munie d'un tel dispositif de commande d'accès
EP0230353A2 (fr) * 1986-01-17 1987-07-29 International Business Machines Corporation Gestion d'un coprocesseur dans un système de traitement de données à mémoire virtuelle et machine virtuelle

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
COMPUTER TECHNOLOGY REVIEW vol. 4, no. 3 , 1984 , LOS ANGELES, CALIFORNIA, USA pages 35 - 38 R. AOUIZERAT 'Shared Memory Devices Can Link PDP-11s Into Load-Balanced Clusters' *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2322209A (en) * 1996-12-23 1998-08-19 Ibm Multi-tasking computer system has shared address space among multiple virtual address spaces
GB2322209B (en) * 1996-12-23 2001-11-21 Ibm Computer system having shared address space among multiple virtual address spaces
US6681239B1 (en) 1996-12-23 2004-01-20 International Business Machines Corporation Computer system having shared address space among multiple virtual address spaces
WO2000063778A1 (fr) * 1999-04-16 2000-10-26 Maxnetwall Computers Oy Dispositif et procede d'amelioration de la securite des informations
US9477516B1 (en) 2015-03-19 2016-10-25 Google Inc. Concurrent in-memory data publication and storage system
CN113580399A (zh) * 2021-07-01 2021-11-02 唐山晶琢科技有限公司 一体化的多线切割机罗拉轴支架

Also Published As

Publication number Publication date
DE69032607D1 (de) 1998-10-08
DE69032607T2 (de) 1999-05-27
CA2009548A1 (fr) 1990-11-17
SG45151A1 (en) 1998-01-16
US5363497A (en) 1994-11-08
ATE170643T1 (de) 1998-09-15
CA2009548C (fr) 1996-07-02
US5144692A (en) 1992-09-01
PT94055A (pt) 1991-11-29
JPH0374756A (ja) 1991-03-29
EP0398695A3 (fr) 1994-02-02
JP2618072B2 (ja) 1997-06-11
BR9002304A (pt) 1991-08-06
EP0398695B1 (fr) 1998-09-02

Similar Documents

Publication Publication Date Title
EP0398695B1 (fr) Mémoire principale physiquement unique, partagée par deux ou plusieurs processeurs exécutant leurs systèmes opérationnels respectifs
EP0400841B1 (fr) Appareil et procédé pour connecter un processeur de données avec un système étranger du traitment des données
EP0398694B1 (fr) Système de traitement de données à tolérance de fautes
EP0398693B1 (fr) Méthode et appareil pour l&#39;addition d&#39;un fonction de traitement des données à un système de traitement de données
US5369749A (en) Method and apparatus for the direct transfer of information between application programs running on distinct processors without utilizing the services of one or both operating systems
EP0405736B1 (fr) Initialisation d&#39;un système de traitement de données à tolérance de fautes
US5369767A (en) Servicing interrupt requests in a data processing system without using the services of an operating system
EP0398697B1 (fr) Communication entre processeurs
KR920008439B1 (ko) 데이타 처리 시스템과 데이타 처리 시스템에 시스템 특성을 추가로 제공하는 방법 및 그 기구

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE CH DE DK ES FR GB GR IT LI LU NL SE

17P Request for examination filed

Effective date: 19901213

PUAL Search report despatched

Free format text: ORIGINAL CODE: 0009013

AK Designated contracting states

Kind code of ref document: A3

Designated state(s): AT BE CH DE DK ES FR GB GR IT LI LU NL SE

17Q First examination report despatched

Effective date: 19960429

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

AK Designated contracting states

Kind code of ref document: B1

Designated state(s): AT BE CH DE DK ES FR GB GR IT LI LU NL SE

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: LI

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 19980902

Ref country code: GR

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 19980902

Ref country code: ES

Free format text: THE PATENT HAS BEEN ANNULLED BY A DECISION OF A NATIONAL AUTHORITY

Effective date: 19980902

Ref country code: CH

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 19980902

Ref country code: AT

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 19980902

REF Corresponds to:

Ref document number: 170643

Country of ref document: AT

Date of ref document: 19980915

Kind code of ref document: T

REG Reference to a national code

Ref country code: CH

Ref legal event code: EP

REF Corresponds to:

Ref document number: 69032607

Country of ref document: DE

Date of ref document: 19981008

REG Reference to a national code

Ref country code: CH

Ref legal event code: NV

Representative=s name: DIPL.-ING. CHRISTIAN HEUSCH C/O INTERNATIONAL BUSI

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SE

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 19981202

Ref country code: DK

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 19981202

ET Fr: translation filed
REG Reference to a national code

Ref country code: CH

Ref legal event code: PL

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GB

Payment date: 19990428

Year of fee payment: 10

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: LU

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 19990516

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: FR

Payment date: 19990518

Year of fee payment: 10

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: DE

Payment date: 19990610

Year of fee payment: 10

PLBE No opposition filed within time limit

Free format text: ORIGINAL CODE: 0009261

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT

26N No opposition filed
PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: GB

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20000516

GBPC Gb: european patent ceased through non-payment of renewal fee

Effective date: 20000516

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: FR

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20010131

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: DE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20010301

REG Reference to a national code

Ref country code: FR

Ref legal event code: ST

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: NL

Payment date: 20010531

Year of fee payment: 12

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: BE

Payment date: 20010620

Year of fee payment: 12

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: BE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20020531

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: NL

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20021201

NLV4 Nl: lapsed or anulled due to non-payment of the annual fee

Effective date: 20021201

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IT

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES;WARNING: LAPSES OF ITALIAN PATENTS WITH EFFECTIVE DATE BEFORE 2007 MAY HAVE OCCURRED AT ANY TIME BEFORE 2007. THE CORRECT EFFECTIVE DATE MAY BE DIFFERENT FROM THE ONE RECORDED.

Effective date: 20050516